[ai] Fan out initial task reviews across all lenses

.agents/skills/continue/SKILL.md

@@ -348,13 +348,17 @@ After it returns, require one of:
 - a clearly reported global hard stop, leaving the task `in-progress` and all
   task-scoped local state recoverable for the next invocation.
 
-A rescope boundary stops this invocation and is reported to the human; it is
+A rescope boundary stops task performance and is not retried, approved, blocked,
-not retried, approved, blocked, or routed as an ordinary discovered follow-up.
+or routed as an ordinary discovered follow-up. The scheduler is the queue
-On a later invocation, a performer that finds the same unresolved boundary
+authority: it independently checks that the proposal contains separately
-returns it immediately without rerunning planning, review, builds, or tests.
+shippable/testable boundaries and a coherent dependency graph before any
-Inbox planning prevents most new oversized tasks; replacing an already-active
+dedicated rescope transaction or human handoff. It never treats the performer's
-task and deciding how to salvage its source is a deliberate queue mutation,
+proposal alone as permission to delete source or rewrite shared task state. On
-not authority the scheduler should infer.
+a later invocation, a performer that finds the same unresolved boundary returns
+it immediately without rerunning planning, review, builds, or tests. Inbox
+planning prevents most new oversized tasks; replacing an already-active task
+and deciding how to salvage its source requires that explicit scheduler-owned
+transaction, not ordinary discovery routing.
 
 Before accepting a canonical test block, read `work/result.md` and
 `work/test.md`. It is genuine only when the verdict is not `TEST_FLAW`, does

.agents/skills/perform-task/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: perform-task
-description: Resolve, start or resume, implement, review, test, and publish exactly one existing ai-tdesktop task by short slug or full dated id, including rare blocked unfinished work. Use when the user invokes $perform-task or /perform-task with a known task name, or when the continue scheduler delegates one selected task. Selects task-specific review specialists and evidence instruments without selecting additional work.
+description: Resolve, start or resume, implement, review, test, and publish exactly one existing ai-tdesktop task by short slug or full dated id, including rare blocked unfinished work. Use when the user invokes $perform-task or /perform-task with a known task name, or when the continue scheduler delegates one selected task. Runs standard review lenses with fast applicability bailouts and selects task-specific domain and evidence instruments without selecting additional work.
 ---
 
 # Perform One AI Task
@@ -124,13 +124,19 @@ must first confirm that the request is one cohesive implementation/review/test
 unit. If it contains independently useful and independently testable product
 boundaries, record `Scope: split-required` and a concrete split proposal before
 source edits, then stop for queue rescoping; do not force the broad request
-through smaller implementation phases and call it one task.
+through smaller implementation phases and call it one task. The independent
+assessment has veto authority over implementation, not authority to create,
+retire, or rewrite tasks. The performer validates and preserves the proposal;
+the checkout scheduler owns any later queue mutation. A direct invocation
+returns the proposal to the human.
 
-For a cohesive task, select one mandatory general review and a falsifiable
+For a cohesive task, use one mandatory general review, all five standard review
-evidence plan. The first general review uses the complete implementation as the
+lenses, and a falsifiable evidence plan. On the initial implementation the
-safety net and chooses specialist reviews only for concrete material questions
+general reviewer and all lenses inspect the task and complete diff without
-that benefit from focused tracing or repository search. Surface presence is a
+seeing one another's findings. A lens may return a compact
-recall prompt, not an automatic specialist. The evidence loop
+`NOT_APPLICABLE` immediately after that scan when it proves the diff affects no
+mechanism it owns; otherwise it reads the relevant changed files and adjacent
+code and returns `CLEAN` or `FINDINGS`. The evidence loop
 may use static readings, commands and artifacts, unit tests, a standalone probe
 or component binary, a Telegram Debug build with logged assertions, an in-app
 overlay, Computer Use, screenshots, or any necessary combination. Do not
@@ -138,9 +144,10 @@ require a portable account, Telegram executable, or desktop unless a selected
 check uses it. Do not weaken a runtime or visual check merely because another
 instrument is cheaper.
 
-The general reviewer examines the complete diff and evidence plan, may require
+The general reviewer examines every changed file in full and the evidence plan,
-a specialist or stronger instrument, and cannot defer a concern to an optional
+may reject an unsupported `NOT_APPLICABLE`, require a named domain specialist
-reviewer. Its approval and every clean specialist result carry forward. A fix
+or stronger instrument, and cannot defer its own concern. Its approval and
+every clean or proved-not-applicable lens result carry forward. A fix
 invalidates only the findings, changed invariants, specialists, validations,
 and evidence checks it actually affects. Review fixes receive a focused general
 delta review plus only those invalidated specialists; they do not restart the

.agents/skills/perform-task/references/phase-prompts.md

@@ -22,9 +22,9 @@ every applicable placeholder: `<TASK>`, `<TASK_ID>`, `<WORK_DIR>`,
 
 ## Orchestration Rules
 
-- When delegation is available, use a fresh subagent for Phase 1 (context and plan), Phase 3, each Phase 4 implementation unit, the initial Phase 6 general review, each selected specialist, each Phase 6 fix, each focused re-review, and any convergence assessment. Do not switch those phases to same-session midstream because of a timeout or missing artifact.
+- When delegation is available, use a fresh subagent for Phase 1 (context and plan), Phase 3, each Phase 4 implementation unit, the initial Phase 6 general review, each of the five standard lens reviews, each Phase 6 fix, each focused re-review, and any convergence assessment. Do not switch those phases to same-session midstream because of a timeout or missing artifact.
-- The initial mandatory general reviewer runs first and alone on the complete diff, and emits a specialist decision table with one concrete material question for every selected lens. Selected specialists are independent and write disjoint reports, so spawn them together when capacity allows, giving them the diff and their question but never the general reviewer's findings. When the host supports continuing the saved general-review agent, send synthesis back to that agent; otherwise use one fresh synthesis agent with the saved general report and specialist reports instead of making it rediscover the whole review.
+- Start the initial mandatory general reviewer and all five lens reviewers together when capacity permits. Under a slot limit, queue the complete set and start the next lens as a slot opens; do not let general review select or prune it. Each lens independently reads the task and complete diff, then either proves `NOT_APPLICABLE` compactly or performs the relevant full review. No reviewer sees another's findings. When the host supports continuing the saved general-review agent, send synthesis back to that agent; otherwise use one fresh synthesis agent with the saved general report and all five lens reports instead of making it rediscover the whole review.
-- After a fix, use the focused re-review prompt: one mandatory general reviewer over the fix and affected invariants, plus only specialists whose recorded question or blocker the fix invalidated. Carry every other approval forward. Do not schedule a lens from surface presence alone.
+- After a fix, use the focused re-review prompt: one mandatory general reviewer over the fix and affected invariants, plus only lenses whose blocker or prior `NOT_APPLICABLE`/`CLEAN` proof the fix invalidated. Carry every other approval forward.
 - Treat delegation as selected only after the first real phase spawn succeeds; tool presence is insufficient. An immediate depth/capacity/policy rejection before phase work selects same-session checklists and is not a delegated retry.
 - Phase 7 runs in the current session on native, non-WSL Windows because it depends on the final local diff and touched-file set. Skip it on WSL and keep files LF/no-BOM there.
 - Write each phase prompt to `<WORK_DIR>/logs/phase-<phase-name>.prompt.md` before execution.
@@ -44,7 +44,7 @@ every applicable placeholder: `<TASK>`, `<TASK_ID>`, `<WORK_DIR>`,
   is the completion signal; there is no polling, no heartbeat-mtime ladder,
   and no stall windows. On return, validate the artifact-based completion
   checks below before treating the phase as done.
-- Spawn the independent leaves of one step — the selected Phase 6 specialists,
+- Spawn the independent leaves of one step — the five initial Phase 6 lenses,
   or assessed-disjoint Phase 4 units —
   as parallel Agent calls in a single message so they run concurrently.
 - If a returned leaf fails its completion check, retry that disposable phase
@@ -58,7 +58,7 @@ every applicable placeholder: `<TASK>`, `<TASK_ID>`, `<WORK_DIR>`,
 - When this session is a top-level `/perform-task`, run each leaf as one
   blocking `spawn_subagent` (`background: false`). The call returning is
   the completion signal; validate the artifact checks below on return.
-- Spawn the independent leaves of one step — the selected Phase 6 specialists,
+- Spawn the independent leaves of one step — the five initial Phase 6 lenses,
   or assessed-disjoint Phase 4
   units — as parallel `spawn_subagent` calls in a single message.
 - When this session is a `/continue` child, do not call `spawn_subagent`.
@@ -128,8 +128,8 @@ Do not restate the full context, plan, diff, or long reasoning in the chat reply
 - Phase 3 is complete only when `plan.md` contains both `Phases:` in the Status section and `Assessed: yes`, records a rejection outcome (`Fast-Path: rejected` or `Approach: rejected`) that sends the performer back to a fresh Phase 1 leaf, or records `Scope: split-required` and has a complete `split-proposal.md` that stops source work for queue rescoping.
 - Phase 4 is complete only when the target phase checkbox changed to checked and the touched-file list matches the owned write set, or the blocker explains any mismatch.
 - Phase 5 is complete only when the build outcome is known and the build checkbox is updated on success.
-- A Phase 6 specialist is complete only when every scheduled specialist wrote `review<R>-<lens>.md` with a `## Verdict:` line and a non-empty `## Checked` section tied to its assigned material question. A report that records no checked surface is incomplete work.
+- An initial Phase 6 lens is complete only when all five `review1-<lens>.md` reports exist with `## Verdict: NOT_APPLICABLE | CLEAN | FINDINGS` and a non-empty `## Checked` section. `NOT_APPLICABLE` must tie its proof to the complete diff; `CLEAN`/`FINDINGS` must name the relevant full files and adjacent surfaces reviewed.
-- Initial Phase 6 general review is complete only when `review1-general.md` and `review1.md` exist with a `## Verdict:` line, a non-empty `## Coverage` section, every standard lens marked `SELECT` or `OMIT` with a valid reason, every selected lens accounted for, and every evidence check reconciled against the actual diff.
+- Initial Phase 6 general review is complete only when `review1-general.md` and `review1.md` exist with a `## Verdict:` line, a non-empty `## Coverage` section, all five lens reports are accepted or an unsupported bailout has been rerun, and every evidence check is reconciled against the actual diff.
 - A focused Phase 6 review is complete only when `review<R>-focused.md` and `review<R>.md` name the fix paths and invariants, account for every carried-forward and invalidated approval, account for every rerun specialist, and reconcile only invalidated evidence checks. A convergence assessment is complete only with one exact disposition from the pipeline and its required bounded next action or split proposal.
 - Phase 6b is complete only when the requested fixes were applied and the post-fix build outcome is known.
 - Phase 3 is additionally incomplete until `test-design.md` exists, covers
@@ -345,7 +345,11 @@ beyond step-level repair. On an approach rejection the performer appends the
 assessor's named simpler direction to the Phase 1 rerun prompt. A third outcome,
 `Scope: split-required`, means the request itself contains several independently
 useful and testable product boundaries; it writes `split-proposal.md` and stops
-before source edits instead of trying another plan for the same task.
+before source edits instead of trying another plan for the same task. This
+independent assessor has veto authority because it is the first phase with the
+exact implementation and evidence plan. It does not create, retire, supersede,
+or rewrite queue tasks; the performer preserves the proposal and returns it to
+the scheduler, which owns any rescope transaction.
 
 ```text
 You are a plan assessment agent. Review and refine an implementation plan.
@@ -388,10 +392,9 @@ Assess the plan:
      beats both growing a mega-module and scattering through one; judge
      whether the boundary does work, not whether it is new.
 5. Expected surfaces: record the surfaces this task is likely to touch and the
-   escalations they would imply, as a recall note for the reviewer. Do not
+   escalations they would imply, as a recall note for the reviewers. Do not
-   select or omit review lenses here — the first general reviewer decides that
+   select or omit standard lenses here — all five independently scan the task
-   with the diff in hand from concrete material questions, not surface presence
+   and complete diff, then decide their own applicability.
-   alone.
 6. Evidence design: map every acceptance criterion and material shipped risk to
    the most direct practical instrument that can detect the negative. Allow
    static readings, commands/artifacts, unit tests, a standalone probe or
@@ -566,25 +569,26 @@ review for another material risk.
 
 For the initial implementation:
 
-1. Run the general reviewer first and alone on the complete diff. It owns the
+1. Launch the general reviewer and all five standard lens reviewers together
-   complete safety review and writes `review1-general.md` with a `SELECT` or
+   when capacity permits. Under a slot limit, queue every lens and start the
-   `OMIT` decision for every standard lens. Every `SELECT` includes one concrete
+   next as a slot opens; none is selected away.
-   material failure question and affected paths/invariants. Surface presence is
+2. Every lens independently reads the task and complete diff. It writes
-   a recall prompt, not a reason by itself.
+   `review1-<lens>.md` with `NOT_APPLICABLE` and a compact proof, or continues
-2. Run selected specialists independently and in parallel. Give each the diff
+   through relevant full files/adjacent code and returns `CLEAN` or `FINDINGS`.
-   and its exact question, never the general reviewer's findings or another
+   It never reads the general reviewer's findings or another lens report.
-   review report. They write `review1-<lens>.md`.
 3. Continue the same general-review agent for synthesis when the host supports
    it. Otherwise use a fresh synthesis agent that starts from
-   `review1-general.md`, reads the specialist reports, and opens only the code
+   `review1-general.md`, reads all five lens reports, and opens only the code
-   needed to confirm their findings. It writes actionable `review1.md`.
+   needed to confirm findings or reject an unsupported bailout. It writes
+   actionable `review1.md`.
 4. `APPROVED` closes review. `NEEDS_CHANGES` runs the fix phase for blocking
    findings only.
 
 For a fix, increment R and do not repeat the initial shape. Run one focused
 general reviewer over the fix result, changed paths/functions, affected callers,
 prior blockers, and invalidated evidence checks. Rerun only a specialist that
-originated a repaired blocker or whose exact recorded question the fix changed.
+originated a repaired blocker or whose prior `NOT_APPLICABLE`/`CLEAN` proof the
+fix invalidated.
 The focused general reviewer confirms those reports and writes `review<R>.md`,
 explicitly carrying all other approvals forward.
 
@@ -605,14 +609,24 @@ You are an independent <LENS> specialist reviewing one Telegram Desktop task.
 You are a leaf and must not delegate.
 
 Read:
-- <WORK_DIR>/context.md
-- <WORK_DIR>/plan.md
-- AGENTS.md and REVIEW.md
 - the task specification
-- the complete task diff and changed files relevant to the assigned question
+- the changed-path manifest and complete task diff, including every hunk
-- the exact material review question and affected paths/invariant
+
+First decide whether this diff affects any mechanism owned by your lens. When
+it does not, write a compact `NOT_APPLICABLE` report tied to exact changed
+paths/hunks and stop. Do not read every changed file in full or search broadly
+just to prove an absent surface. Uncertainty means applicable; small size,
+`documentation only`, time pressure, or low estimated severity do not prove
+non-applicability.
+
+When applicable, also read:
+- <WORK_DIR>/context.md and plan.md
+- AGENTS.md and REVIEW.md
+- every relevant changed file in full
+- adjacent callers, owners, consumers, or repository precedents needed for this
+  lens
 - for R > 1, only the preceding actionable finding and fix result assigned to
-  this specialist
+  this lens
 
 Do not read another reviewer's findings. Do not search or read
 `<WORK_DIR>/review*` or phase-review logs beyond the exact files listed above;
@@ -621,8 +635,8 @@ report is exposed accidentally, disclose it and stop so only this specialist
 can be rerun cleanly.
 
 Review only this task's diff under your assigned angle. Search outside the diff
-when your angle requires repository or call-site context, but do not report
+only after the lens is applicable and when call-site or repository context is
-pre-existing problems as task findings.
+needed. Do not report pre-existing problems as task findings.
 
 A finding is BLOCKING only when it names a concrete wrong result, crash, race,
 security or data-safety failure, material performance regression, violated
@@ -635,14 +649,14 @@ Write <WORK_DIR>/review<R>-<LENS>.md:
 ## Lens: <LENS> — iteration <R>
 
 ## Checked
-<Every relevant changed surface examined for the assigned question and what
+<For NOT_APPLICABLE: the complete-diff proof that no owned mechanism changed.
-was established.>
+For CLEAN/FINDINGS: relevant full files and adjacent surfaces examined.>
 
 ## Findings
 <For each: title, file/line, Severity: BLOCKING | NON_BLOCKING, concrete
 failure, and specific fix. Omit when empty.>
 
-## Verdict: CLEAN | FINDINGS
+## Verdict: NOT_APPLICABLE | CLEAN | FINDINGS
 ~~~
 
 ### Specialist: lifetime
@@ -744,31 +758,29 @@ Read:
 - on pass 1, the complete task diff and every changed file in full, plus
   adjacent callers, consumers, generated/build declarations, and repository
   precedents needed to judge integration
-- on pass 2, review1-general.md, every selected review1-<lens>.md, and only the
+- on pass 2, review1-general.md, all five review1-<lens>.md reports, and only
-  code needed to confirm/drop a specialist finding or resolve a contradiction
+  the code needed to confirm/drop a finding, validate a NOT_APPLICABLE proof,
+  or resolve a contradiction
 
 Independently review correctness, completeness, edge/error paths, unintended
 regressions, integration, proportionality, repository conventions, and the
 evidence design. Do not defer anything to a specialist.
 
-On pass 1 no specialist has run yet. Besides your own review, emit a specialist
+On pass 1 the five standard lenses are running independently. Complete your own
-decision table. Mark every optional lens — lifetime, reuse, structure,
+review without predicting or selecting their verdicts. Name any extra domain
-performance, security — exactly `SELECT` or `OMIT`; silence is invalid.
+specialist a material risk needs. Write `review1-general.md` with your complete
+independent Checked, evidence reconciliation, findings, and pass-1 status, but
+do not write `review1.md` yet. Then return so the performer can call you back for
+synthesis after every standard lens report exists.
 
-`SELECT` only when the actual diff presents a concrete material failure question
+On pass 2, account for all five lenses. Confirm every finding against the code
-for which focused tracing, call-site search, or domain expertise adds useful
+and drop it when the concrete failure does not hold. Accept `NOT_APPLICABLE`
-independent confidence. State the question, affected paths/invariant, and the
+only when its complete-diff proof establishes that no owned mechanism changed.
-failure it could expose. `OMIT` when the surface is absent or narrow enough that
+If that proof contradicts a hunk or is merely a severity estimate, write
-your complete review already established its relevant invariant with no open
+`Incomplete lens: <name> — <owned mechanism>` and return without an
-specialist question; state that proof. Surface presence alone, `low risk`, time
+implementation verdict; the performer reruns only that lens as applicable and
-pressure, or `documentation only` is not a decision. Name any extra domain
+calls you back. Do not redo the complete general review; carry pass 1 findings
-specialist a material risk needs. Then return: the performer runs selected
+and coverage forward and open only code needed for synthesis.
-specialists and calls you back for pass 2.
-
-On pass 2, confirm every specialist finding against the code; drop it when the
-concrete failure does not hold. Do not redo the complete general review; carry
-pass 1 findings and coverage forward and open only the code needed for
-synthesis or a newly contradicted invariant.
 
 Reconcile test-design.md against the actual diff:
 - every acceptance criterion and material new risk has a check;
@@ -784,10 +796,9 @@ Reconcile test-design.md against the actual diff:
 - selected instrument prerequisites are explicit, and unavailable unselected
   instruments are not treated as blockers.
 
-If pass 2 exposes a concrete material question no selected lens covered, write
+If pass 2 exposes a material domain outside the five standard lenses, write
-"Missing specialist: <name> — <question>" and return without an implementation
+`Missing specialist: <domain> — <question>` and return without an implementation
-verdict. The performer runs only that specialist, then calls you back. Do not
+verdict. The performer runs only that domain specialist, then calls you back.
-turn a generic surface into a late specialist request.
 
 Classify findings:
 - BLOCKING: concrete wrong behavior, crash, race, security/data-safety failure,
@@ -796,14 +807,13 @@ Classify findings:
 - NON_BLOCKING: optional wording, preference, speculative cleanup, or polish.
   Preserve it in Dropped/Notes, but do not ask the fix agent to implement it.
 
-Write <WORK_DIR>/review1-general.md with Checked, the specialist decision table,
+On pass 2, update <WORK_DIR>/review1-general.md with lens confirmation and the
-specialist confirmation,
+final general verdict. Then write <WORK_DIR>/review1.md:
-evidence reconciliation, findings and verdict. Then write
-<WORK_DIR>/review1.md:
 
 ## Code Review — Initial
 ## Coverage
-<general coverage plus every SELECT/OMIT reason and material question>
+<general coverage plus all five NOT_APPLICABLE/CLEAN/FINDINGS results and any
+domain specialist>
 ## Evidence reconciliation
 <each planned check confirmed or changed>
 ## Verdict: APPROVED | NEEDS_CHANGES | MISSING_SPECIALIST
@@ -827,7 +837,7 @@ Read:
 - <WORK_DIR>/review<R-1>.md and the fix result;
 - the fixed paths, containing functions/types, and affected callers;
 - test-design.md entries the fix reports invalidated;
-- only the specialist reports explicitly rerun for this fix.
+- only the lens or domain-specialist reports explicitly rerun for this fix.
 
 Do not restart the complete-diff review. Verify each repaired blocker, inspect
 the fix for regressions in its actual data/control/lifetime boundary, and check
@@ -836,11 +846,11 @@ record `CARRIED` with why the fix did not touch its exact code or invariant, or
 `INVALIDATED` with the concrete changed question. Presence of a broad surface is
 not invalidation.
 
-Rerun a specialist only when it originated a repaired blocker, its exact prior
+Rerun a lens only when it originated a repaired blocker, the fix invalidated
-question changed, or this fix introduced a new concrete material question that
+its prior `NOT_APPLICABLE`/`CLEAN` proof, or the fix introduced a new mechanism
-focused expertise must answer. If one is needed and has not run, write
+owned by that lens. If one is needed and has not run, write
-`Missing specialist: <name> — <question>` and return; the performer runs only
+`Missing lens: <name> — <invalidated proof or new mechanism>` and return; the
-that specialist and calls you back.
+performer runs only that lens and calls you back.
 
 Reconcile only evidence checks whose changed surface, oracle, fixture, or
 expected result the fix invalidated. Carry all other checks forward.
@@ -918,9 +928,9 @@ pre-review validation once after review approval. Use build-lock recovery when
 applicable.
 
 Report exact touched paths, repaired findings, changed functions/invariants,
-which prior specialist question (if any) the fix invalidated and why, and which
+which prior lens `NOT_APPLICABLE`/`CLEAN` result (if any) the fix invalidated
-validation/evidence checks it invalidated. Surface presence alone is not
+and why, and which validation/evidence checks it invalidated. Surface presence
-invalidation.
+alone is not invalidation.
 If no blocking finding can be acted on inside the owned write set, change
 nothing and report that boundary.
 ~~~
@@ -1089,7 +1099,7 @@ For each phase:
 4. Save `<WORK_DIR>/logs/phase-<phase-name>.result.md` with `STATUS:`, `ARTIFACTS:`,
    `TOUCHED:`, `BLOCKER:`, and `NOTES:` fields.
 
-For review iterations, include the iteration and selected lens in the file name, for example:
+For review iterations, include the iteration and lens in the file name, for example:
 - `phase-1-context-plan.prompt.md`
 - `phase-6a-review-1-general.prompt.md`
 - `phase-6a-review-1-general.result.md`

.agents/skills/perform-task/references/pipeline.md

@@ -140,8 +140,8 @@ work/project.proposed.md       # project tasks only
 work/visual.md                 # layout tasks only
 work/plan.md
 work/split-proposal.md         # only when assessment rejects intrinsic scope
-work/review1-general.md        # mandatory complete-diff review and selection
+work/review1-general.md        # mandatory complete-file review
-work/review1-<specialist>.md   # specialists selected for material questions
+work/review1-<lens>.md         # all five standard lens applicability/reviews
 work/review1.md                # canonical first-review verdict
 work/review<R>-focused.md      # mandatory focused general re-review, R > 1
 work/review<R>-<specialist>.md # only specialists invalidated by the fix
@@ -203,7 +203,7 @@ prompts, plus the host-specific orchestration rules.
   policy rejects that spawn before work begins, execute the same prompt
   checklists in the performer. This is a supported mode, not degraded failure.
 - In nested mode, use a fresh leaf for context-and-plan, assessment, each
-  implementation unit, selected specialist reviews, general review,
+  implementation unit, initial standard-lens reviews, general review,
   review-fix, and evidence authoring. Every leaf must be told not to delegate and
   never to commit.
 - Small-task fast path: the performer may run the context-and-plan checklist
@@ -276,6 +276,10 @@ Run sequentially:
    `Approach: rejected` outcome that reruns Phase 1 as a fresh leaf — an
    approach rejection with the assessor's simpler direction as added input —
    or the terminal pre-edit `Scope: split-required` planning boundary.
+   This assessor has authority to veto source implementation because it has the
+   first exact implementation/evidence plan. It cannot mutate task state beyond
+   its local artifacts or create replacement tasks; the scheduler independently
+   validates and owns rescoping.
 3. **Implement.** Run one leaf per assessed plan phase. Before each edit,
    update `work/owned-paths.txt`. A leaf edits only its owned paths and its
    phase status; it does not commit.
@@ -287,11 +291,13 @@ Run sequentially:
    resource, force its documented regeneration. Apply exact-path Windows
    cleanup and build-lock recovery only to a command that writes the configured
    build tree.
-5. **Review.** Run the adaptive review/fix loop from the phase prompts. The
+5. **Review.** Run the adaptive review/fix loop from the phase prompts. Launch
-   first independent general reviewer always examines the complete diff,
+   the first independent general reviewer and all five standard lens reviewers
-   adjacent integration, task contract, and `work/test-design.md`, and selects
+   against the task and complete diff without sharing findings. The general
-   specialists for concrete material questions. It confirms or drops their
+   reviewer examines every changed file in full; each lens either proves
-   findings and writes the single `review1.md` the fix phase implements. Only
+   `NOT_APPLICABLE` from the complete diff or performs its relevant focused
+   review. General synthesis confirms or drops findings, rejects unsupported
+   bailouts, and writes the single `review1.md` the fix phase implements. Only
    material blocking findings cause a fix. After a fix, preserve every prior
    approval that the fix did not invalidate: run one focused general review of
    the fix and affected invariants, plus only the specialists that originated a
@@ -352,34 +358,38 @@ Evidence plan:
 - Evidence: <planned durable output>
 ```
 
-Assessment does not choose reviewers. The first general reviewer does that with
+Assessment does not choose among the standard reviewers. `Expected surfaces` is
-the diff in hand, and `Expected surfaces` is a recall checklist rather than an
+a recall checklist for their initial applicability scan. Assessment rejects
-automatic decision. Assessment rejects ceremony as well as under-testing: it
+ceremony as well as under-testing: it removes evidence checks that cannot be
-removes checks that cannot be affected by the task.
+affected by the task.
 
-### Review selection
+### Initial review fanout
 
-Selection happens with the implementation in hand. Assessment records expected
+The initial review always includes one independent **general** reviewer and all
-surfaces and escalation triggers, but a task-text prediction neither schedules
+five standard lenses. Start them together when capacity permits; under a slot
-nor omits a reviewer.
+limit, keep the complete set queued and start the next lens as soon as a slot
+opens. The general reviewer and lenses do not read one another's findings.
 
-The independent **general** review runs first, alone, exactly once for the
+The general reviewer reads every changed file in full and owns correctness,
-initial implementation. It reads every changed file in full and owns
+completeness, adjacent integration, unintended regressions, proportionality,
-correctness, completeness, adjacent integration, unintended regressions,
+repository conventions, and evidence adequacy. It cannot defer a concern to a
-proportionality, repository conventions, and evidence adequacy. It cannot defer
+specialist. This is the mandatory safety net.
-a concern to a specialist. This complete-diff pass is the mandatory safety net.
 
-Its first output is a **specialist decision table**. Account for every lens
+Each lens first reads the task specification and complete task diff — every
-below with `SELECT` or `OMIT`; silence is invalid. Select a specialist only when
+hunk, not necessarily every changed file in full — and chooses one result:
-the diff creates a concrete material failure question that focused tracing,
-call-site search, or domain expertise is likely to answer better than the
-general pass alone. Record that question, the affected paths or invariant, and
-what a failure would do. Omit it when the surface is absent, or when the surface
-is narrow and the general reviewer has already established the relevant
-invariant with no unresolved specialist question. `Low risk`, time pressure,
-and `documentation only` are not sufficient reasons by themselves.
 
-The lens surfaces are recall prompts, not automatic retention clauses:
+- `NOT_APPLICABLE` — the diff affects no mechanism owned by the lens. Give a
+  short, concrete proof tied to the changed paths/hunks, then stop without broad
+  repository searches or reading irrelevant files in full.
+- `CLEAN` — the lens applies; read the relevant changed files in full and the
+  adjacent callers, owners, consumers or precedents needed to decide it, and
+  report no material finding.
+- `FINDINGS` — perform the same applicable review and report concrete material
+  findings.
+
+Uncertainty means the lens is applicable. `Low risk`, small size, time pressure,
+and `documentation only` are not proofs of non-applicability. The standard lens
+surfaces are:
 
 - **lifetime** — object and resource ownership, callbacks, re-entrancy,
   destruction order, threads, concurrency, races, synchronization,
@@ -394,24 +404,21 @@ The lens surfaces are recall prompts, not automatic retention clauses:
   untrusted input, command or subprocess construction, filesystem boundaries,
   downloads, network trust, and destructive behavior.
 
+Lens source search is rooted at `SOURCE_ROOT`; a reviewer must not search or read
+`WORK_DIR/review*` or phase-review logs. Accidental exposure is reported and
+invalidates only that lens, not already independent work.
+
 The general reviewer may add a named domain specialist when a material risk
-such as ABI portability or persistence migration fits none of the lenses.
+such as ABI portability or persistence migration fits none of the five lenses.
-
+Reuse the same general reviewer for synthesis when the host supports continuing
-Run selected specialists independently and in parallel. Give each the task
+that agent; otherwise a fresh synthesis reviewer starts from the saved general
-diff, its exact material review question, and source paths, but never the
+report and lens reports, then reads only code needed to confirm or drop their
-general reviewer's findings or another review report. Specialist source search
+findings. Synthesis rejects a `NOT_APPLICABLE` whose proof contradicts the diff
-is rooted at `SOURCE_ROOT`; it must not search or read `WORK_DIR/review*` or
+and sends only that lens back for a full applicable review. It writes the sole
-phase-review logs beyond the explicit input list. Accidental exposure is
+`review1.md` verdict and accounts for all five lens results. Wording, style, or
-reported and invalidates only that specialist, not already independent work.
+optional cleanup that does not cause wrong behavior, unsafe use, material
-
+maintenance cost, or a repository-rule violation is non-blocking and never
-A specialist reports only material findings with a concrete failure. Reuse the
+starts a fix cycle.
-same general reviewer for synthesis when the host supports continuing that
-agent; otherwise a fresh synthesis reviewer starts from the saved general
-report and specialist reports, then reads only code needed to confirm or drop
-their findings. It writes the sole `review1.md` verdict and accounts for every
-lens decision. Wording, style, or optional cleanup that does not cause wrong
-behavior, unsafe use, material maintenance cost, or a repository-rule violation
-is non-blocking and never starts a fix cycle.
 
 ### Focused re-review and convergence
 
@@ -422,16 +429,16 @@ invariants, and invalidated validations/evidence. After the fix:
 
 1. run one mandatory focused general review over those edits, the containing
    functions and affected callers, and the prior blocking findings;
-2. rerun only a specialist that originated a repaired blocker or whose recorded
+2. rerun only a lens that originated a repaired blocker or whose prior
-   material question the fix actually changed;
+   `NOT_APPLICABLE`/`CLEAN` proof the fix actually invalidated;
 3. reconcile only evidence checks invalidated by the fix; and
 4. carry every other general, specialist, validation, and evidence approval
    forward explicitly.
 
-The focused general reviewer owns synthesis. It may add a specialist only for a
+The focused general reviewer owns synthesis. It reruns a previously inapplicable
-new concrete material question introduced by the fix. Merely touching an async,
+lens only when the fix introduces one of its owned mechanisms. Merely touching
-storage, build, security, or other broad surface does not invalidate every lens
+an async, storage, build, security, or other broad surface does not invalidate
-that could describe it. It reads the complete task diff only when the fix has
+every lens that could describe it. It reads the complete task diff only when the fix has
 made the previous review boundary unreliable, in which case it triggers the
 convergence assessment rather than silently restarting the initial review.
 
@@ -820,8 +827,10 @@ delays finishing the work actually in hand.
   convergence assessment are planning boundaries, not task `Block` verdicts
   and not permission to keep retrying. Preserve `split-proposal.md`, all
   validated source recovery, and the task's `in-progress` state; return the
-  proposal to the scheduler or human and stop automatic performance until the
+  proposal to the scheduler (or human for a direct invocation) and stop
-  queue is deliberately replaced. A resumed performer that sees the same
+  automatic performance until the queue is deliberately replaced. Assessment
+  owns the stop; only the scheduler owns task creation, dependency rewrites,
+  and superseding the original. A resumed performer that sees the same
   unresolved boundary reports it immediately without rerunning context,
   implementation, review, or builds.
 - Source lineage has a strict timing boundary. Before Phase 1, a missing

.agents/skills/process-inbox/SKILL.md

@@ -129,6 +129,15 @@ owns and the direct evidence that can approve it without first implementing a
 sibling. If that sentence needs several independent outcomes or several
 unrelated instruments, split again.
 
+This is the first scope gate, not an irrevocable ruling. Inbox planning uses the
+request plus light source inspection and deliberately does not construct the
+implementation plan. The later independent perform-task assessment sees exact
+files, APIs, phases, ownership boundaries, and evidence design; it may veto the
+single-task shape when that richer proof exposes independently shippable and
+testable boundaries. That veto does not mean task sizing is based on elapsed
+time or diff length, and it does not authorize the performer to mutate the
+queue itself.
+
 Project slugs are unique across `projects/` and `projects/archive/`. When a
 request belongs to an archived project, restore it before routing to it: