arcasilesgroup Verify
Verifies merge-readiness claims with evidence instead of assumptions, dispatching deterministic and acceptance specialists over changed files. Use it to check code, prove tests pass, scan for security issues, or run the 8-dimension --release GO/NO-GO gate.
Quick start
/ai-verify # normal: deterministic + acceptance (LLM)
/ai-verify --full # one agent per specialist
/ai-verify quality # deterministic quality scan only
/ai-verify platform # 2-specialist aggregate verdict (post-W3)
/ai-verify --release [version] # 8-dimension release gate (GO|CONDITIONAL GO|NO-GO)
Workflow
Principles applied: §10.5 TDD (tests and gates ARE the evidence — run the command, read the exit code, never assume). Evidence before claims. This SKILL.md owns the user-facing contract; verifier agent files provide specialist lenses and must not redefine mode semantics.
- Step 0 — load contexts — read
.ai-engineering/manifest.ymlproviders.stacks; apply.ai-engineering/overrides/<stack>/conventions.mdper stack. - Dependency preflight — verify
handlers/verify.mdplus the.claude/agents/verifier-*.mdfiles the selected mode needs exist (normal/--fullneed deterministic + acceptance; single modes need only their specialist). STOP and report the exact missing path — never improvise. - Run IRRV protocol — per claim: identify command -> run -> capture output + exit code -> classify CONFIRMED (exit 0 + expected) or REFUTED.
- Dispatch specialists via the Agent tool (never read inline). Report every finding by its original specialist lens.
Dispatch threshold
Dispatch the ai-verify agent for any merge-readiness check, scan, or evidence-backed claim over ≥ 1 changed file. Each specialist runs in its own context window via the Agent tool. .claude/agents/ai-verify.md is the orchestrator handle; the procedural contract lives here.
Specialist Roster (post-W3: 2)
Spec-140 W3: verifier-governance + verifier-feature merged into verifier-acceptance; verifier-architecture heuristics moved to /ai-advise (advisory) and the standalone verifier deleted.
| Specialist | Agent File | Lens |
|---|---|---|
deterministic |
verifier-deterministic.md |
Security, quality, dependencies, tests (tool-driven; always runs first) |
acceptance |
verifier-acceptance.md |
Spec coverage, acceptance criteria, governance compliance, ownership boundaries, gate enforcement (LLM) |
Modes
| Mode | What runs |
|---|---|
normal (implicit) |
deterministic, then acceptance (single LLM macro) |
--full |
same 2 specialists, dispatched in parallel after deterministic |
quality / security |
deterministic agent only (one scan slice) |
acceptance / governance / feature |
acceptance specialist only (governance/feature aliases preserved for muscle memory) |
platform |
aggregate verdict over deterministic + acceptance |
--release [version] |
8-dimension release-readiness gate (D-127-10). Stack-detected (Python/JS/Rust/Go); aggregates coverage (≥ manifest threshold), security (gitleaks + semgrep + pip-audit, zero crit/high), tests (100% pass), lint (zero unfixable), dependency vulns (zero CVEs unless risk-accepted in decision-store.json), types (zero errors), documentation coherence (CHANGELOG current), packaging integrity (build clean). Verdict GO (all PASS) / CONDITIONAL GO (PASS with risk acceptances) / NO-GO (>= 1 blocker). Closure path printed for NO-GO. |
Both profiles run the same two specialists — grouping differs (single macro vs. parallel), not coverage. Deterministic always runs first and feeds the acceptance judgment. See handlers/verify.md for orchestration.
Output Contract
Every scan mode emits: score / verdict (PASS/WARN/FAIL) / profile / specialist table / findings grouped by specialist / gate check.
| Mode | Blocker if… | Critical if… |
|---|---|---|
| deterministic | Any secret detected, any test failure | Coverage < 80%, critical lint |
| acceptance | Spec goal missing, integrity FAIL, suppression added | Acceptance criterion unmet, compliance FAIL, count drift |
| platform | Any blocker in ANY mode | Score < 60 |
Verification Checklist (before claiming DONE)
- Every acceptance criterion verified with a command.
- All tests pass (exact count reported).
- Lint/format clean (zero warnings).
- No secrets in staged files.
- Coverage maintained or improved (exact % reported).
- No forbidden words ("should work") in the completion report.
Common Mistakes
- Assuming
--fulladds specialist coverage instead of changing decomposition. - Reporting a specialist as skipped instead of
not applicable. - Ignoring warnings when the exit code is 0.
- Reading specialist agent files inline instead of dispatching via the Agent tool.
Examples
Example — pre-merge platform sweep
User: "is this branch ready to merge?"
/ai-verify platform
Dispatches deterministic + acceptance in parallel (post-W3 roster of 2), aggregates findings, scores against the gate, returns PASS / WARN / FAIL with evidence per finding.
Integration
Called by: /ai-build (post-task), /ai-autopilot (Phase 5), user directly. Dispatches: verifier-deterministic, verifier-acceptance agents. Read-only: never modifies code. See also: /ai-review (narrative review), /ai-advise (advisory architecture lens), /ai-reliability-eval, /ai-security (deep CVE/SBOM only), /ai-governance (compliance, risk acceptance).
Inline fallback: Agent-tool dispatch is the primary path. On a harness with no subagent/Agent-tool primitive, execute this skill by reading the needed .claude/agents/verifier-*.md specialist file(s) and running their steps inline, in-context, sequentially — inline-sequential is the floor, not the default.
$ARGUMENTS