Recently Written · git

ai-config

One canonical AI agent config (rules, skills, MCP servers) rendered into Claude Code, Codex, Copilot CLI and Antigravity by acsync

git clone https://github.com/equwal/ai-config

Log | Files | Refs


skills/verification-before-completion/SKILL.md (3646 bytes)

1 ---
2 name: verification-before-completion
3 description: Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always
4 ---
5 
6 # Verification Before Completion
7 
8 ## Overview
9 
10 **Core principle:** Evidence before claims, always.
11 
12 **Violating the letter of this rule is violating the spirit of this rule.**
13 
14 ## The Iron Law
15 
16 ```
17 NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE
18 ```
19 
20 If you haven't run the verification command in this message, you cannot claim it passes.
21 
22 ## The Gate Function
23 
24 ```
25 BEFORE claiming any status or expressing satisfaction:
26 
27 1. IDENTIFY: What command proves this claim?
28 2. RUN: Execute the FULL command (fresh, complete)
29 3. READ: Full output, check exit code, count failures
30 4. VERIFY: Does output confirm the claim?
31    - If NO: State actual status with evidence
32    - If YES: State claim WITH evidence
33 5. ONLY THEN: Make the claim
34 
35 Skip any step = lying, not verifying
36 ```
37 
38 ## Common Failures
39 
40 | Claim | Requires | Not Sufficient |
41 |-------|----------|----------------|
42 | Tests pass | Test command output: 0 failures | Previous run, "should pass" |
43 | Linter clean | Linter output: 0 errors | Partial check, extrapolation |
44 | Build succeeds | Build command: exit 0 | Linter passing, logs look good |
45 | Bug fixed | Test original symptom: passes | Code changed, assumed fixed |
46 | Regression test works | Red-green cycle verified | Test passes once |
47 | Agent completed | VCS diff shows changes | Agent reports "success" |
48 | Requirements met | Line-by-line checklist | Tests passing |
49 
50 ## Red Flags - STOP
51 
52 - Using "should", "probably", "seems to"
53 - Expressing satisfaction before verification ("Great!", "Perfect!", "Done!", etc.)
54 - About to commit/push/PR without verification
55 - Trusting agent success reports
56 - Relying on partial verification
57 - Thinking "just this once"
58 - Tired and wanting work over
59 - **ANY wording implying success without having run verification**
60 
61 ## Rationalization Prevention
62 
63 | Excuse | Reality |
64 |--------|---------|
65 | "Should work now" | RUN the verification |
66 | "I'm confident" | Confidence ≠ evidence |
67 | "Just this once" | No exceptions |
68 | "Linter passed" | Linter ≠ compiler |
69 | "Agent said success" | Verify independently |
70 | "I'm tired" | Exhaustion ≠ excuse |
71 | "Partial check is enough" | Partial proves nothing |
72 | "Different words so rule doesn't apply" | Spirit over letter |
73 
74 ## Key Patterns
75 
76 **Tests:**
77 ```
78 ✅ [Run test command] [See: 34/34 pass] "All tests pass"
79 ❌ "Should pass now" / "Looks correct"
80 ```
81 
82 **Regression tests (TDD Red-Green):**
83 ```
84 ✅ Write → Run (pass) → Revert fix → Run (MUST FAIL) → Restore → Run (pass)
85 ❌ "I've written a regression test" (without red-green verification)
86 ```
87 
88 **Build:**
89 ```
90 ✅ [Run build] [See: exit 0] "Build passes"
91 ❌ "Linter passed" (linter doesn't check compilation)
92 ```
93 
94 **Requirements:**
95 ```
96 ✅ Re-read plan → Create checklist → Verify each → Report gaps or completion
97 ❌ "Tests pass, phase complete"
98 ```
99 
100 **Agent delegation:**
101 ```
102 ✅ Agent reports success → Check VCS diff → Verify changes → Report actual state
103 ❌ Trust agent report
104 ```
105 
106 ## When To Apply
107 
108 **ALWAYS before:**
109 - ANY variation of success/completion claims
110 - ANY expression of satisfaction
111 - ANY positive statement about work state
112 - Committing, PR creation, task completion
113 - Moving to next task
114 - Delegating to agents
115 
116 **Rule applies to:**
117 - Exact phrases
118 - Paraphrases and synonyms
119 - Implications of success
120 - ANY communication suggesting completion/correctness