Menu
Decide whether a change is safe to merge — runs `act gate`, the deterministic verdict (MERGE / REVIEW / BLOCK / UNKNOWN) over every changed function. Use before merging a branch or approving a PR.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Audit whether an AI agent edit to a file is safe — composes secret-surface, taint-flow, change-impact, and API-surface analysis into an agent-edit safety report. Use before letting an agent modify a sensitive or high-blast-radius file.
Execute structural decompositions based on an architectural analysis report. Use when breaking circular dependencies, splitting god classes, reducing coupling between modules, extracting interfaces at seam boundaries, or executing any large-scale structural change identified in an architectural analysis. Requires a prior architecture audit (the architecture-audit skill) that produced project-map.md and a run report.
Use when asked to audit architecture, produce a comprehensive architectural overview, get the full structural picture of a codebase, assess overall health and porting readiness, find module boundaries, circular dependencies, coupling hotspots, dead code, or code patterns. Depth 3 — full audit with hypothesis-driven investigation. Optionally enriches the structural audit with runtime and git evidence (coverage, churn, co-change coupling, ownership) when those overlays are available. Replaces the architectural-analysis and architecture-audit-plus skills.
Use when looking for extraction candidates, planning a module split, analyzing module boundaries, or before decomposing a large component. Also use when asked "where should I split this?" or "find natural module boundaries". Depth 2 — investigate. Produces extraction candidates with cut costs, layer violations, interface width assessment, orphan types, and a recommended decomposition sequence.
Use before modifying a file or symbol, or when asked "what breaks if I change X?". Depth 0 — fast, no follow-up queries, returns immediately. Two modes: inline (no artifacts, default for agent-initiated use) and artifact (writes to docs/act/, triggered for human-initiated requests or Critical risk verdicts).
Traverse large repositories efficiently using act's query tools. Use when exploring unfamiliar code, mapping dependencies, understanding API surfaces, analyzing side effects before modifying functions, or following call chains across files. Avoids reading entire files by querying only the structure needed.
| name | safe-to-merge |
| description | Decide whether a change is safe to merge — runs `act gate`, the deterministic verdict (MERGE / REVIEW / BLOCK / UNKNOWN) over every changed function. Use before merging a branch or approving a PR. |
Delegate to the gate tool (CLI twin: act gate). The verdict rules live
in the engine — one source of truth; this skill describes how to run it and
read the result, it does not re-specify the rules.
| Scope | Call |
|---|---|
| Working tree vs HEAD (default) | gate with no params / act gate |
| PR branch vs its base | gate with base_ref = the base (e.g. origin/main) / act gate --base-ref origin/main |
CLI exit codes: 0 MERGE, 1 BLOCK, 2 REVIEW, 3 UNKNOWN, 4 execution error — CI scripts branch on exact codes (they are identifiers, not a severity scale).
Discovers every changed function from the git diff (rename-aware; same-file
exact-body renames are paired as signature changes), then composes the
verification trio per function: verify_diff_semantics (hunk classes —
guard-condition changes count as behavior), verify_test_impact (does any
test reach it — syntactic call-graph floor with import resolution),
verify_side_effects (effect delta, dropped cleanup). Broadly: format- or
signature-only changes with test reach merge; tested behavior changes ask for
review; untested behavior/signature changes and dropped cleanup block; anything
the grammar cannot model degrades to UNKNOWN with modeled_kinds quoted.
Changed test code is exempt from the reach rule. The authoritative matrix is in
act-analysis's gate module — read the per-function reason field; it cites
the rule that fired.
These are AST/heuristic analyses, not a proof. UNKNOWN means "not verified" — never present it as MERGE; the engine enforces this (UNKNOWN outranks MERGE and REVIEW in the overall verdict). Per-grammar degradation is real: grammars without branch-CFG modeling cannot rule out behavior changes in body edits and return UNKNOWN with the evidence quoted.
Engineer (the verify trio's tier; named Pro before V2). Tier-blocked runs and premium-language files report UNKNOWN with the blocking dimension named — report that, do not guess a verdict.
verdict is the overall worst case over functions[]. Each record carries:
classification, tested + impacted_tests, effects_added/effects_removed,
dropped_cleanup, modeled_kinds, per-function verdict, and reason.
skipped[] lists changed files that did not gate (no grammar) — mention them
in your summary so coverage is honest.
