dos-witness-claim
Skill49 starsChanged 15 days agoApache-2.0View on GitHub
Verify subagent or worker results before folding them into a synthesis. Use when another agent claims it shipped, created, found, or changed something and your next step would otherwise trust its return string.
In this repository151 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- llms-full.txtllms.txt
- llms.txtllms.txt
- agent-readinessSkill
- appeal-scoreSkill
- bottleneck-mapSkill
- ci-repairSkill
- claim-repro-scoreSkill
- clear-out-wipSkill
- commit-cleanSkill
- conflation-scoreSkill
- cross-validateSkill
- curate-clusterSkill
- debt-cleanSkill
- debt-orchestratorSkill
- disambiguation-scoreSkill
- dojo-rsi-scoreSkill
- dos-dispatch-loopSkill
- dos-dispatchSkill
- dos-next-upSkill
- dos-replanSkill
- field-borrowSkill
- fleet-waveSkill
- git-subagent-syncSkill
- goalSkill
- guard-rsi-scoreSkill
- harness-creatorSkill
- harness-gardenSkill
- industry-scoreSkill
- issue-orchestratorSkill
- issue-queueSkill
- issue-triageSkill
- lightgap-scoreSkill
- memory-compactSkill
- milestone-scoreSkill
- modularizeSkill
- negframe-scoreSkill
- opencode-waveSkill
- operator-heaviness-scoreSkill
- persona-scoreSkill
- phased-planSkill
- plan-auditSkill
- quality-scoreSkill
- question-loopSkill
- refresh-cachedoc-numbersSkill
- refresh-readmeSkill
- releaseSkill
- resume-watchdog-auditSkill
- run-it-all-nightSkill
- score-2xSkill
- scorecardSkill
- scout-loopSkill
- signal-firstSkill
- skill-lifecycleSkill
- skill-overlapSkill
- skill-scoreSkill
- slop-scoreSkill
- sota-checkSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 2 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 18
Never interpolate a worker's return string directly into a synthesis as ground truth.
- Very specific: line 28
Read paths, lanes, and stamp from the result.
- Distinctive: line 34
If you have a worker transcript path, verify the result is not a harness-authored death before reading it as a worker result:
- Very specific: line 48
If verify-result is not available in the installed DOS version, log the gap and continue to effect witnessing.
- Very specific: line 83
Treat OK / diff-witnessed as foldable shape evidence.
- Very specific: line 97
If the read-back confirms the effect, fold it as CONFIRMED.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
dos-witness-claim - fold witnessed effects, not narration
A worker result is a claim, not a fact. Before using it as input to another prompt, report, plan, or commit, witness the claimed effect from a surface the worker did not author. Fold only confirmed effects; route unconfirmed results separately and count them in the denominator.
Load-Bearing Rule
Never interpolate a worker's return string directly into a synthesis as ground truth. Extract the checkable effect, gather a non-agent-authored witness, and fold only the confirmed bucket.
Step 0 - Discover The Workspace
dos doctor --workspace . --jsonRead paths, lanes, and stamp from the result. If a claim names a
(plan, phase), dos verify applies this workspace's stamp grammar; do not
grep commit subjects yourself.
Step 1 - Check Worker Terminal State
If you have a worker transcript path, verify the result is not a harness-authored death before reading it as a worker result:
dos verify-result --workspace . --transcript <agent-transcript.jsonl>Route by exit code:
0: terminal result is usable for claim extraction; still witness the effect.3: worker is dead or synthetic. Do not fold its return string; re-dispatch or surface the failed unit.2: contract error. Fix the transcript wiring; do not treat it as healthy.
If verify-result is not available in the installed DOS version, log the gap and
continue to effect witnessing. The gap is not a pass.
Step 2 - Extract The Checkable Claim
Classify each result into one of these claim types:
| Claim type | Example | Witness |
| Git phase | "PLAN / PHASE shipped" | dos verify over git ancestry and stamp grammar |
| Commit claim | "commit <sha> did X" | dos commit-audit <sha> |
| File/effect claim | "created file X", "sent id Y" | fresh read-back, OS exit code, API GET, or state diff |
| No checkable claim | "done", "looks good" | no fold; surface as NO_CLAIM |
Abstain rather than inventing identifiers from prose. A result without a named effect is not confirmed.
Step 3 - Witness Git-Backed Claims
For a (plan, phase) claim:
dos verify --workspace . <PLAN> <PHASE> --jsonFold only when the exit code is shipped and the JSON carries a positive source
such as registry or grep-subject. Preserve the source rung in the synthesis;
grep-subject is weaker than a registry row.
For a commit claim:
dos commit-audit --workspace . <sha>Treat OK / diff-witnessed as foldable shape evidence. Treat
CLAIM_UNWITNESSED as residual review work, not as success.
Step 4 - Witness Non-Git Effects
There is no generic dos verify-effect CLI verb. For created files, DB rows,
messages, deploys, or external side effects, gather a fresh read-back from a
surface the worker did not author:
- file claim: fresh filesystem read/stat of the exact path;
- API claim: GET the object by id;
- DB claim: read the row by key;
- command claim: run the relevant command and check its exit/status output.
If the read-back confirms the effect, fold it as CONFIRMED. If it contradicts
the claim, route it as REFUTED. If no accountable read-back exists, route it as
UNWITNESSED; never fold it as success.
Step 5 - Return A Partition, Not A Story
Return results in four buckets:
CONFIRMED: safe to fold into the next prompt or report.REFUTED/NOT_SHIPPED/DEAD: do not fold; re-dispatch or surface.UNWITNESSED: do not fold; identify the missing witness.NO_CLAIM: do not fold; ask for a checkable effect if needed.
Carry coverage forward:
confirmed M of N declared results; refuted R; unwitnessed U; no-claim CDownstream synthesis must see the coverage. Do not let 4/7 confirmed become
7/7 returned.
Anti-Patterns
- Folding
${result}directly because it is non-empty. - Asking a model whether a worker's own output looks complete.
- Counting only returned strings in the denominator.
- Treating
UNWITNESSEDorNO_CLAIMas probably successful. - Grepping commit subjects yourself instead of calling
dos verify.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

