Skip to main content
Certificate of Conformance #002. One of the factory’s own quality instruments, pulled through Certify so the tools that gauge everything else carry lot numbers too. validate-work-order is the intake linter: it refuses a malformed work order before any production run starts.

​
The card

A tool is a deterministic instrument. This one is stdlib-only, binds no model at build or run time, and its verdict is a function of its input, not a probability.

​
Bill of materials

The fixture suite is a constituent, not a test fixture on the side: the tool’s proof of work is --self-test, and that self-test is meaningless without the planted-defect fixtures it proves against. So the fixtures are content-hashed into the BOM alongside the script.
  • workorders/validate-work-order.py, sha256 bba57996d1dac61af20682d2b0195466190379094f2e3285bad64e3375e8a425
  • workorders/fixtures/bad-missing-family.md, sha256 c83fb6e892e651539f93a9751ca31e51bad8bb2903541f35f031d633efc818b9
  • workorders/fixtures/bad-missing-principal.md, sha256 3fc797095654de3f3ac8fee9b44664579b9904a3b427d519febed2c03dec2654
  • workorders/fixtures/bad-missing-target-cost.md, sha256 746b27c8e9e5e9d5190214296c862d74e6303b47be6f4b9549deb549224151ee
  • workorders/fixtures/bad-missing-why.md, sha256 d39f93d5389c03d1ff21ac19967ca8343f13e15c61d09ef74ebd8a9b7cab2706
  • workorders/fixtures/bad-script-shaped-mission.md, sha256 7b68be773ed411458c60e83dfee8df3dd7b9c008effcccfb241344ed92d0b1b3
  • workorders/fixtures/bad-untestable-end-state.md, sha256 624f487f38bbff6fb9bd9281b3460cabcb27ede6a5684a15cf8c4f19e5d4a07c
  • workorders/fixtures/good-agent.md, sha256 fbb8ea87d5fb1883a6f33e66e946596583ce91e2e56a5279e73ec4e5b7c66e66
  • workorders/fixtures/good-skill.md, sha256 27537fcccf300761c386170f65b08b3df0976dc3a1733ec9083e3ac7d5e6e34b
  • workorders/fixtures/good-tool.md, sha256 24ff3a8039f044abab2a1246f7611061ef311f37404baae85b2dcedc8e29c76f
  • Part: none
  • Model: none (deterministic). No model binding at build or run time.

​
Evidence

  • Eval: --self-test fixture battery, 9 of 9. Six planted-defect fixtures were each rejected for exactly its planted reason, and three good fixtures were accepted clean. No separate tool-family eval suite exists; the self-test is the family’s eval, recorded as a finding.
  • Proof of work: the raw --self-test output, verdict PASS in both directions.
  • Judge: none. The judge harness is not built yet (a published gap). The probabilistic layer was manual: a second agent’s independent review.
  • Traces: the assembly record (PR #101, the Station 0 intake gate) and the certification run (WO-A8, issue #132).

​
Provenance

  • Work order: WO-A8, issue #132.
  • Assembling agent: original assembly and certification both by worker sessions.
  • Trial runs: none required. Supervised trial runs are the agent family’s gauge; a tool’s proof is its deterministic fixture suite, run in both directions.

​
Liability

  • Owner of record: the operator of record.
  • Escalation: the consuming session or continuous-integration job opens an issue on the factory’s public repository, and the operator of record takes the call.
  • Recall: kill switch. Revert the tool and its fixtures to the last certified revision, mark the certificate recalled, find root cause by Five Whys, fix the die (the linter or the work-order schema) upstream, re-run the Verify gates, and re-issue.

​
Expiry

  • Re-certify by: 2026-08-10 (30 days from issue).
  • Triggers: any BOM hash mismatch (the tool or its fixtures changed, so the certified unit no longer exists); a runtime dependency change (the interpreter version the factory runs it under); a change to the work-order schema this tool compiles (its verdict would no longer mean the same thing); or 3 production failures. A model swap does not apply: no model binding.

​
First-pass yield

Clean. Both this tool and its sibling (#003) opened green at Verify, with no defect verdicts to adjudicate. The honest note is on the other side: the self-test is the eval suite for now, so its coverage is only as wide as the fixtures planted, and that limit is recorded in the evidence rather than hidden by the passing result.

​
Self-audit