diff --git a/pi/core/skills/agent-self-evaluation/SKILL.md b/pi/core/skills/agent-self-evaluation/SKILL.md index 4e241380a..ff9b7d4ea 100644 --- a/pi/core/skills/agent-self-evaluation/SKILL.md +++ b/pi/core/skills/agent-self-evaluation/SKILL.md @@ -15,7 +15,16 @@ After completing a complex task, the agent pauses to rate its own output against - After a debugging session that involved 3+ attempts - After producing a design document, architecture decision, or written analysis - When the user asks "how good was that?" or "rate yourself" -- At the end of any session Stop hook (if configured — see `references/hook-integration.md`) +- At the end of any session Stop hook (if configured — see the [hook integration guide](references/hook-integration.md)) + +## Supporting Files + +- [Evaluation criteria](references/evaluation-criteria.md) — concrete score anchors for distinguishing adjacent ratings +- [Hook integration](references/hook-integration.md) — opt-in Stop and PostToolUse reminders +- [High-score example](examples/high-score-example.md) — a fully verified evaluation report +- [Low-score example](examples/low-score-example.md) — an evaluation that identifies substantive gaps +- [Report template](templates/evaluation-report.md) — copy-ready scorecard and scoring triggers +- [Evaluator script](scripts/evaluate.py) — optional heuristic first pass for file or piped output ## Core Concepts @@ -69,7 +78,7 @@ Do NOT average the scores in your head first and then work backwards. Score each ### Step 3: Produce the Evaluation Report -Use the template from `templates/evaluation-report.md`. The report must include: +Use the [evaluation report template](templates/evaluation-report.md). The report must include: ``` - One-line summary diff --git a/skills/agent-self-evaluation/SKILL.md b/skills/agent-self-evaluation/SKILL.md index 4e241380a..ff9b7d4ea 100644 --- a/skills/agent-self-evaluation/SKILL.md +++ b/skills/agent-self-evaluation/SKILL.md @@ -15,7 +15,16 @@ After completing a complex task, the agent pauses to rate its own output against - After a debugging session that involved 3+ attempts - After producing a design document, architecture decision, or written analysis - When the user asks "how good was that?" or "rate yourself" -- At the end of any session Stop hook (if configured — see `references/hook-integration.md`) +- At the end of any session Stop hook (if configured — see the [hook integration guide](references/hook-integration.md)) + +## Supporting Files + +- [Evaluation criteria](references/evaluation-criteria.md) — concrete score anchors for distinguishing adjacent ratings +- [Hook integration](references/hook-integration.md) — opt-in Stop and PostToolUse reminders +- [High-score example](examples/high-score-example.md) — a fully verified evaluation report +- [Low-score example](examples/low-score-example.md) — an evaluation that identifies substantive gaps +- [Report template](templates/evaluation-report.md) — copy-ready scorecard and scoring triggers +- [Evaluator script](scripts/evaluate.py) — optional heuristic first pass for file or piped output ## Core Concepts @@ -69,7 +78,7 @@ Do NOT average the scores in your head first and then work backwards. Score each ### Step 3: Produce the Evaluation Report -Use the template from `templates/evaluation-report.md`. The report must include: +Use the [evaluation report template](templates/evaluation-report.md). The report must include: ``` - One-line summary