| name | sprint-contract-observability |
| description | Add process observability for agent work. Use when a task needs sprint contracts, evaluator rubrics, task traces, acceptance evidence, or structured feedback between generator and reviewer roles. |
Sprint Contract Observability
Purpose
Make agent process intent observable through contracts, rubrics, traces, and evidence-backed evaluator feedback.
Inspect First
- current task plan, specs, feature list, observability docs, review rubric, validation scripts, browser/runtime debugging docs, and recent QA feedback
Procedure
-
Create a sprint contract.
- Use
docs/observability/sprint-contract.md.
- Define scope, exclusions, touched surfaces, verification standards, runtime signals, and handoff expectations.
-
Create an evaluator rubric.
- Use
docs/observability/evaluator-rubric.md.
- Score correctness, product behavior, architecture, tests, accessibility, observability, and maintainability when relevant.
-
Record task trace when useful.
- Use
docs/observability/task-trace.jsonl for task events, checks, failures, fixes, and evidence.
- Keep entries compact and machine-readable.
- Use the canonical trajectory event shape: version, timestamp, task id, task type, phase, action, files read, files changed, commands run, validation, risk surfaces, intervention, outcome, and notes.
- Route trajectory classification and advisory risk scoring to
$trajectory-risk-supervisor.
-
Add advisory validation.
- Create or recommend
scripts/check-sprint-contract.
- Use
scripts/check_sprint_contract.py from this plugin.
-
Coordinate runtime observability.
- Use
observability-harness for logs, diagnostics, correlation IDs, and health checks.
- Use
chrome-devtools-runtime-debugger for live browser evidence.
- Use
agent-reviewer for findings-first evaluation.
- Use
$subagent-review-orchestrator when explicit reviewer, browser QA, docs research, or trajectory auditor subagents would improve evaluation.
Validation
- Run
./scripts/check-sprint-contract when present.
- Confirm scope, exclusions, verification standards, rubric, and trace are consistent.
- Confirm evaluator feedback cites evidence instead of subjective approval.
Completion Criteria
- The task has observable process intent and structured evaluation criteria.
- Future sessions can see why work was accepted, rejected, or deferred.