completion-gate

Name: completion-gate
Author: Rune-kit

80stars21forks

How to install completion-gate?

rune-kit/rune/completion-gate

$curl -o .claude/agents/completion-gate.md https://raw.githubusercontent.com/Rune-kit/rune/HEAD/agents/completion-gate.md

Installs into the current project.

Summary

Lie detector for agent claims — validates every completion claim has actual evidence. Default-FAIL mindset. Use as final gate before merge/commit.

Files · 1

View on GitHub

agents/completion-gate.md

1	# completion-gate
2
3	## Purpose
4
5	The lie detector for agent claims. Validates that what an agent says it did actually happened — with evidence. Catches the #1 failure mode in AI coding: claiming completion without proof.
6
7	<HARD-GATE>
8	Every claim requires evidence. No evidence = UNCONFIRMED = BLOCK.
9	"I ran the tests and they pass" without stdout = UNCONFIRMED.
10	"I fixed the bug" without before/after diff = UNCONFIRMED.
11	"Build succeeds" without build output = UNCONFIRMED.
12	</HARD-GATE>
13
14	## Triggers
15
16	- Called by `cook` in Phase 5d (quality gate)
17	- Called by `team` before merging stream results
18	- Called by any skill that reports "done" to an orchestrator
19	- Auto-trigger: when agent says "done", "complete", "fixed", "passing"
20
21	## Calls (outbound)
22
23	None — pure validator. Reads evidence, produces verdict.
24
25	## Called By (inbound)
26
27	- `cook` (L1): Phase 5d — validate completion claims before commit
28	- `team` (L1): validate cook reports from parallel streams
29
30	## Execution
31
32	### Step 1 — Collect Claims
33
34	Parse the agent's output for completion claims. Common claim patterns:
35
36	```
37	CLAIM PATTERNS:
38	"tests pass" / "all tests passing" / "test suite green"
39	"build succeeds" / "build complete" / "compiles clean"
40	"no lint errors" / "lint clean"
41	"fixed" / "resolved" / "bug is gone"
42	"implemented" / "feature complete" / "done"
43	"no security issues" / "sentinel passed"
44	```
45
46	Extract each claim as: `{ claim: string, source_skill: string }`
47
48	### Step 1a — Type Each Claim (Claim Discipline)
49	<MUST-READ path="references/claim-discipline.md" trigger="always — before matching evidence"/>
50
51	Before hunting for evidence, type the claim by the grammar it was written in. Hallucination is an unverified claim wearing the grammar of an observation — the grammar is the tell, and it is readable in the sentence itself.
52
53	\| Type \| Meaning \| Grammar it may wear \|
54	\|------\|---------\|---------------------\|
55	\| OBSERVED \| Seen this session: ran it, read it, measured it \| "X is / does / returns …" \|
56	\| DERIVED \| Follows from OBSERVED facts via a statable mechanism \| "X should / will / implies …" + the why \|
57	\| PRIOR \| Training knowledge, may be stale \| "X is typically … / was, as of …" \|
58	\| ASSUMED \| Unverified and required by the conclusion \| "I am assuming X — if wrong, then …" \|
59
60	This changes what the gate is looking for in Step 2:
61
62	- OBSERVED → demands an evidence artifact. No artifact = FAIL. This is the existing gate.
63	- DERIVED → demands the mechanism be stated, and its OBSERVED inputs to be present.
64	- PRIOR / ASSUMED → not a failure. A claim honestly delivered as assumed is the correct output when the check was not run. Record it as an open item; never score it as a lie.
65
66	<HARD-GATE>
67	A hedge is not a defect. Do NOT fail a claim for being marked ASSUMED or PRIOR — fail it for
68	wearing OBSERVED grammar with nothing behind it. Treating honest uncertainty as a failure
69	teaches the next agent to delete its hedges, which is the exact behavior this gate exists to
70	catch.
71	</HARD-GATE>
72
73	Claims are promoted only by tools — checking a PRIOR makes it OBSERVED. Restating it more confidently does not. Confidence that grew from effort, repetition or fluent prose resets to the last evidence-backed level.
74
75	### Step 1b — Stub Detection (Existence Theater Check)
76
77	Before checking claims, scan all files created/modified in this workflow for stubs:
78
79	```
80	Grep for stub patterns in new/modified files:
81	- "Placeholder" \| "TODO" \| "Not implemented" \| "NotImplementedError"
82	- Functions with body: only `return null` / `return {}` / `pass` / `throw`
83	- Components returning only a single div with no logic
84	```
85
86	If ANY stub detected:
87	- Add synthetic claim: "implemented [filename]" → CONTRADICTED (file is a stub)
88	- This catches agents that create files but don't implement them
89
90	### Step 1c — Self-Validation Check
91
92	If the skill that just ran has a `## Self-Validation` section, extract its checklist and treat each item as an implicit claim:
93
94	```
95	For each Self-Validation check in the skill's SKILL.md:
96	1. Read the check (e.g., "at least one assertion per test")
97	2. Look for evidence in tool output that this check was satisfied
98	3. If evidence found → add as CONFIRMED claim
99	4. If no evidence → add as UNCONFIRMED claim ("Self-Validation: [check] — no evidence")
100	```
101
102	Why: Self-Validation catches domain-specific quality issues that generic claim matching (Step 2) cannot detect. A test skill knows "no assertions = useless test" but completion-gate doesn't — unless the skill's Self-Validation tells it to check.
103
104	<HARD-GATE

People who liked this also viewed

6 picks

Type

completion-gate

80stars21forks

View on GitHub

How to install completion-gate?

rune-kit/rune/completion-gate

$curl -o .claude/agents/completion-gate.md https://raw.githubusercontent.com/Rune-kit/rune/HEAD/agents/completion-gate.md

Installs into the current project.

Summary

Lie detector for agent claims — validates every completion claim has actual evidence. Default-FAIL mindset. Use as final gate before merge/commit.

Files · 1

View on GitHub

agents/completion-gate.md

1	# completion-gate
2
3	## Purpose
4
5	The lie detector for agent claims. Validates that what an agent says it did actually happened — with evidence. Catches the #1 failure mode in AI coding: claiming completion without proof.
6
7	<HARD-GATE>
8	Every claim requires evidence. No evidence = UNCONFIRMED = BLOCK.
9	"I ran the tests and they pass" without stdout = UNCONFIRMED.
10	"I fixed the bug" without before/after diff = UNCONFIRMED.
11	"Build succeeds" without build output = UNCONFIRMED.
12	</HARD-GATE>
13
14	## Triggers
15
16	- Called by `cook` in Phase 5d (quality gate)
17	- Called by `team` before merging stream results
18	- Called by any skill that reports "done" to an orchestrator
19	- Auto-trigger: when agent says "done", "complete", "fixed", "passing"
20
21	## Calls (outbound)
22
23	None — pure validator. Reads evidence, produces verdict.
24
25	## Called By (inbound)
26
27	- `cook` (L1): Phase 5d — validate completion claims before commit
28	- `team` (L1): validate cook reports from parallel streams
29
30	## Execution
31
32	### Step 1 — Collect Claims
33
34	Parse the agent's output for completion claims. Common claim patterns:
35
36	```
37	CLAIM PATTERNS:
38	"tests pass" / "all tests passing" / "test suite green"
39	"build succeeds" / "build complete" / "compiles clean"
40	"no lint errors" / "lint clean"
41	"fixed" / "resolved" / "bug is gone"
42	"implemented" / "feature complete" / "done"
43	"no security issues" / "sentinel passed"
44	```
45
46	Extract each claim as: `{ claim: string, source_skill: string }`
47
48	### Step 1a — Type Each Claim (Claim Discipline)
49	<MUST-READ path="references/claim-discipline.md" trigger="always — before matching evidence"/>
50
51	Before hunting for evidence, type the claim by the grammar it was written in. Hallucination is an unverified claim wearing the grammar of an observation — the grammar is the tell, and it is readable in the sentence itself.
52
53	\| Type \| Meaning \| Grammar it may wear \|
54	\|------\|---------\|---------------------\|
55	\| OBSERVED \| Seen this session: ran it, read it, measured it \| "X is / does / returns …" \|
56	\| DERIVED \| Follows from OBSERVED facts via a statable mechanism \| "X should / will / implies …" + the why \|
57	\| PRIOR \| Training knowledge, may be stale \| "X is typically … / was, as of …" \|
58	\| ASSUMED \| Unverified and required by the conclusion \| "I am assuming X — if wrong, then …" \|
59
60	This changes what the gate is looking for in Step 2:
61
62	- OBSERVED → demands an evidence artifact. No artifact = FAIL. This is the existing gate.
63	- DERIVED → demands the mechanism be stated, and its OBSERVED inputs to be present.
64	- PRIOR / ASSUMED → not a failure. A claim honestly delivered as assumed is the correct output when the check was not run. Record it as an open item; never score it as a lie.
65
66	<HARD-GATE>
67	A hedge is not a defect. Do NOT fail a claim for being marked ASSUMED or PRIOR — fail it for
68	wearing OBSERVED grammar with nothing behind it. Treating honest uncertainty as a failure
69	teaches the next agent to delete its hedges, which is the exact behavior this gate exists to
70	catch.
71	</HARD-GATE>
72
73	Claims are promoted only by tools — checking a PRIOR makes it OBSERVED. Restating it more confidently does not. Confidence that grew from effort, repetition or fluent prose resets to the last evidence-backed level.
74
75	### Step 1b — Stub Detection (Existence Theater Check)
76
77	Before checking claims, scan all files created/modified in this workflow for stubs:
78
79	```
80	Grep for stub patterns in new/modified files:
81	- "Placeholder" \| "TODO" \| "Not implemented" \| "NotImplementedError"
82	- Functions with body: only `return null` / `return {}` / `pass` / `throw`
83	- Components returning only a single div with no logic
84	```
85
86	If ANY stub detected:
87	- Add synthetic claim: "implemented [filename]" → CONTRADICTED (file is a stub)
88	- This catches agents that create files but don't implement them
89
90	### Step 1c — Self-Validation Check
91
92	If the skill that just ran has a `## Self-Validation` section, extract its checklist and treat each item as an implicit claim:
93
94	```
95	For each Self-Validation check in the skill's SKILL.md:
96	1. Read the check (e.g., "at least one assertion per test")
97	2. Look for evidence in tool output that this check was satisfied
98	3. If evidence found → add as CONFIRMED claim
99	4. If no evidence → add as UNCONFIRMED claim ("Self-Validation: [check] — no evidence")
100	```
101
102	Why: Self-Validation catches domain-specific quality issues that generic claim matching (Step 2) cannot detect. A test skill knows "no assertions = useless test" but completion-gate doesn't — unless the skill's Self-Validation tells it to check.
103
104	<HARD-GATE

People who liked this also viewed

6 picks

Type