version: "1.0.1" name: shipyard-verification description: Use when about to claim work is complete, fixed, or passing — before committing, creating PRs, or moving to the next task. Also use when you catch yourself using words like "should", "probably", or "seems to" about work state, or when expressing satisfaction ("Great!", "Done!") before running verification commands. Evidence before assertions, always.

<!-- TOKEN BUDGET: 180 lines / ~540 tokens -->

Verification Before Completion

<activation>

Activation Triggers

• About to claim work is "done", "complete", "fixed", or "passing"
• About to commit, create a PR, or merge
• Before any success assertion -- must have fresh evidence first
• Using words like "should", "probably", "seems to" about work state

Natural Language Triggers

• "is this done", "verify this", "check my work", "are we done", "does this work"

</activation>

Overview

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

Core principle: Evidence before claims, always.

Violating the letter of this rule is violating the spirit of this rule.

<instructions>

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

Teammate Mode

When SHIPYARD_IS_TEAMMATE=true:

• Run verification locally and report results via task metadata (not STATE.json updates)
• Do not update STATE.json — only the lead agent writes project state
• TaskCompleted hook enforces evidence — the hook checks for verification artifacts before allowing task completion

In solo mode, this section has no effect.

</instructions>

<rules>

Red Flags - STOP

• Using "should", "probably", "seems to"
• Expressing satisfaction before verification ("Great!", "Perfect!", "Done!", etc.)
• About to commit/push/PR without verification
• Trusting agent success reports
• Relying on partial verification
• Thinking "just this once"
• Tired and wanting work over
• ANY wording implying success without having run verification

Rationalization Prevention

When To Apply

ALWAYS before:

• ANY variation of success/completion claims
• ANY expression of satisfaction
• ANY positive statement about work state
• Committing, PR creation, task completion
• Moving to next task
• Delegating to agents

Rule applies to:

• Exact phrases
• Paraphrases and synonyms
• Implications of success
• ANY communication suggesting completion/correctness

</rules>

<examples>

Verification Examples

Good: Claim backed by evidence

Agent: I'll run the tests to verify.
 
$ npm test
> 34 passing (1.2s)
> 0 failing
 
All 34 tests pass. The refactored auth module is working correctly.

Why it works: The agent ran the command, read the output, and THEN made the claim. The evidence is inline and verifiable.

Bad: Assertion without evidence

Agent: I've updated the auth module to fix the token validation bug.
The tests should pass now. Let me move on to the next task.

Why it fails: "should pass" is not evidence. No command was run. The agent is guessing based on the code change, not verifying the outcome.

Good: Honest report of partial success

Agent: Running the full test suite:
 
$ pytest tests/
> 28 passed, 2 failed
 
28 of 30 tests pass. Two failures remain in test_auth_refresh.py:
- test_expired_token_refresh: expects 401 but gets 200
- test_concurrent_refresh: race condition in token store
 
The original bug (token validation) is fixed, but these two
related failures need investigation before this is complete.

Bad: Trusting an agent's self-report

Agent: The builder agent reported success. Phase complete!

Why it fails: Agent self-reports are not verification. Check the VCS diff and run the tests yourself.

</examples>

Why This Matters

From failure analysis: your human partner said "I don't believe you" -- trust broken. Undefined functions shipped. Missing requirements shipped. Time wasted on false completion then redirect then rework.

No shortcuts for verification. Run the command. Read the output. THEN claim the result. This is non-negotiable.