context-doctor

bunx smithers-orchestrator workflow run context-doctor \
  --input '{"contract":"{\"goal\":\"Add CSP headers\",\"acceptanceCriteria\":[{\"text\":\"headers present\",\"blocking\":true,\"verification\":\"curl -I\"}],\"inputs\":[{\"name\":\"repo\",\"required\":true,\"source\":\"arg\"}],\"sideEffects\":[{\"name\":\"deploy\",\"approval\":true}],\"outputSpec\":\"patch\",\"reportSpec\":\"summary\"}"}'

​

Stages

1. check: deterministic (no agent), JSON.parse the contract and run the structural checks, returning { issues, summary, score }.
2. advise: an agent reads the diagnosis and writes one { check, fix } pair per non-ok issue (check is the check id being resolved), ordered error → warning → info.

​

Checks

• hasGoal: goal is non-empty.
• hasOutputSpec: outputSpec (or output) describes the produced artifact.
• hasAcceptanceCriteria: at least one acceptance criterion is listed.
• allBlockingCriteriaHaveVerification: every blocking criterion names a verification. Plain-string criteria warn: a string can’t declare blocking or verification, so use {text, blocking, verification} objects.
• allRequiredInputsHaveSource: every required input declares a source.
• allSideEffectsHaveApproval: every sideEffects entry is gated by an approval. Plain-string entries count as unguarded; use {name, approval} objects.
• reportSpecExists: reportSpec describes how results are reported.

​

Inputs

​

Monitor it

bunx smithers-orchestrator workflow run context-doctor --input '{"contract":"…"}' -d   # detach
bunx smithers-orchestrator ps                                                          # active / paused / recent
bunx smithers-orchestrator logs RUN_ID -f                                             # follow events
bunx smithers-orchestrator output RUN_ID --node check                                 # inspect the deterministic diagnosis