Identify Issues in Session Transcript

You are an expert AI quality reviewer analyzing session transcripts to surface actionable issues in a LearningAgent's behavior.

Arguments

$ARGUMENTS is the path to the session log folder (e.g., .deepwork/tmp/agent_sessions/<session_id>/<agent_id>/).

Context

Agent used: !cat $ARGUMENTS/agent_used 2>/dev/null || echo "unknown"

Last learning timestamp (empty if never learned): !cat $ARGUMENTS/learning_last_performed_timestamp 2>/dev/null

Existing issue files (avoid duplicates): !ls $ARGUMENTS/*.issue.yml 2>/dev/null || echo "(none)"

Additional identification guidelines: !learning_agents/scripts/cat_agent_guideline.sh $ARGUMENTS issue_identification

Session log folder structure: !cat learning_agents/doc/learning_log_folder_structure.md 2>/dev/null

Procedure

Step 1: Read the Transcript

Read $ARGUMENTS/conversation_transcript.jsonl. It's JSONL — focus on type: "assistant" messages and type: "tool_result" entries. If learning_last_performed_timestamp exists (shown above), skip lines before that timestamp.

Step 2: Identify Issues

Look for these categories of problems:

1. Incorrect outputs: Wrong answers, broken code, invalid configurations
2. Knowledge gaps: The agent didn't know something it should have
3. Missed context: Information was available but the agent failed to use it
4. Poor judgment: Questionable decisions or suboptimal approaches
5. Pattern failures: Repeated errors suggesting a systemic issue

Skip trivial issues (minor formatting, environmental failures, issues already covered by existing learnings or issue files listed above).

Step 3: Report Each Issue

For each issue, invoke the report-issue skill:

Skill learning-agents:report-issue $ARGUMENTS "<one-sentence description>" "<timestamp of relevant line(s)>"

Step 4: Clean Up if No Issues

If zero issues were found, delete the needs_learning_as_of_timestamp file from the session folder:

rm $ARGUMENTS/needs_learning_as_of_timestamp

This marks the session as fully processed so that the investigate and incorporate steps can be skipped.

Step 5: Summary

## Session Issue Summary

**Session**: <session_id>
**Agent**: <agent_used>
**Issues found**: <count>

| # | Category | Brief description |
|---|----------|-------------------|
| 1 | <category> | <one sentence> |

(or: "No actionable issues found. No follow-up needed — session marked as processed.")

Guardrails

• Do NOT investigate root causes — that is the next step's job
• Do NOT modify the agent's knowledge base
• Do NOT create duplicate issues for the same problem
• Focus on actionable issues that can lead to concrete improvements