gregoryho/arc-verifying

arc-verifying

Core Principle

Claiming work is complete without verification is dishonesty, not efficiency.

Boundary

arc-verifying owns producing fresh evidence for completion claims. It does not own authoring spec artifacts (that is arc-refining) and it does not own reconciling spec/code drift after implementation (that is the optional, separate, future arc-syncing-spec workflow — never folded into the SessionStart bootstrap or the arc-using router). Spec/code drift checks may quote verification evidence as input, but verification itself is not a spec-sync skill.

The Iron Law

NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE

If you haven't run the verification command in THIS message, you cannot claim it passes.

The Gate Function

BEFORE claiming any status or expressing satisfaction:

1. IDENTIFY: What command proves this claim?
2. RUN: Execute the FULL command (fresh, complete)
3. READ: Full output, check exit code, count failures
4. VERIFY: Does output confirm the claim?
   • If NO: State actual status with evidence
   • If YES: State claim WITH evidence
5. ONLY THEN: Make the claim

Skip any step = lying, not verifying

Common Failures

| Claim | Requires | NOT Sufficient | |-------|----------|----------------| | Tests pass | Test command output: 0 failures | Previous run, "should pass" | | Build succeeds | Build command: exit 0 | Linter passing | | Bug fixed | Test original symptom: passes | Code changed, assumed fixed | | Agent completed | VCS diff shows changes | Agent reports "success" | | Requirements met | Line-by-line checklist | Tests passing |

Regression Tests (Red/Green)

If claiming a bug is fixed, require a true regression check:

1. Run failing test (RED)
2. Apply fix
3. Run test again (GREEN)

Skipping RED means you don't know the test proves anything.

Requirements Verification

If claiming requirements are met:

1. Re-read the requirements
2. Make a checklist
3. Verify each item with evidence
4. Report any gaps explicitly

When Verification Cannot Run

Core principle: Cannot verify ≠ skip verification. Must inform user and choose alternative.

Flow

digraph cannot_verify {
    "Cannot run verification?" [shape=diamond];
    "Inform user immediately" [shape=box];
    "User chooses" [shape=diamond];
    "Fix the blocker" [shape=box];
    "Add debug output" [shape=box];
    "User reports result" [shape=box];

    "Cannot run verification?" -> "Inform user immediately" [label="yes"];
    "Inform user immediately" -> "User chooses";
    "User chooses" -> "Fix the blocker" [label="fix env"];
    "User chooses" -> "Add debug output" [label="manual test"];
    "Add debug output" -> "User reports result";
}

Handling

| Situation | Action | |-----------|--------| | Build fails | Fix build first, then verify | | Cannot run Simulator/Emulator | Ask user: fix blocker OR add debug print | | Requires manual UI testing | Describe expected behavior, ask user to verify |

Rationalizations

| Excuse | Reality | |--------|---------| | "Should work now" | Assumption ≠ verification | | "I changed it, should be fine" | Changed ≠ verified | | "Continue for now, verify later" | Cannot verify = stop here |

Red Flags - Cannot Verify

• Cannot verify but don't inform user
• Multiple changes without any verification
• Assuming "should work"

Red Flags - STOP

• Using "should", "probably", "seems to"
• Expressing satisfaction BEFORE verification ("Great!", "Perfect!", "Done!")
• About to commit/push/PR without verification
• Trusting agent success reports
• Relying on partial verification
• Assuming linter success implies build/test success
• Feeling tired and wanting it over
• Using different wording to imply success without evidence

Rationalization Prevention

| Excuse | Reality | |--------|---------| | "Should work now" | RUN the verification | | "I'm confident" | Confidence ≠ evidence | | "Just this once" | No exceptions | | "Agent said success" | Verify independently | | "Partial check is enough" | Partial proves nothing | | "Too simple to verify" | Complexity irrelevant | | "I already ran it earlier" | Run it again, now | | "The logs look fine" | Logs ≠ verification |

Key Patterns

Tests:

✅ [Run test command] [See: 34/34 pass] "All tests pass"
❌ "Should pass now" / "Looks correct"

Build:

✅ [Run build] [See: exit 0] "Build passes"
❌ "Linter passed" (linter doesn't check compilation)

Agent delegation:

✅ Agent reports success → Check VCS diff → Verify changes → Report actual state
❌ Trust agent report

Requirements:

✅ Re-read plan → Create checklist → Verify each → Report gaps or completion
❌ "Tests pass, phase complete"

Integration

Discoverable from: arc-using when a task is approaching a completion claim and verification guidance would help.

Also embedded in:

• arc-finishing / arc-finishing-epic — verify tests before offering merge options
• arc-tdd — Verify RED / Verify GREEN steps
• Spec reviewer / Quality reviewer — read actual code, run tests

Invoke this skill explicitly before finishing. Embedded verification in other skills is an additional layer, not a replacement.