From learning-agents
Analyzes LearningAgent session transcripts to identify mistakes, knowledge gaps, underperformance, and creates YAML issue files for problems found.
npx claudepluginhub unsupervisedcom/deepwork --plugin learning-agentsThis skill uses the workspace's default tool permissions.
You are an expert AI quality reviewer analyzing session transcripts to surface actionable issues in a LearningAgent's behavior.
Investigates identified issues in LearningAgent sessions by analyzing transcripts, determining root causes like knowledge gaps or missing docs, and updating YAML issue files with reports.
Analyzes Orchestrator/Worker/Reviewer transcripts for issues to detect structural problems, protocol deviations, anti-patterns, and propose fixes.
Analyzes Claude Code JSONL transcripts to detect anti-patterns, tool misuse, user frustration signals, and workflow patterns using DuckDB SQL, 10 dimensions, and PM4Py mining.
Share bugs, ideas, or general feedback.
You are an expert AI quality reviewer analyzing session transcripts to surface actionable issues in a LearningAgent's behavior.
$ARGUMENTS is the path to the session log folder (e.g., .deepwork/tmp/agent_sessions/<session_id>/<agent_id>/).
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
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.
Look for these categories of problems:
Skip trivial issues (minor formatting, environmental failures, issues already covered by existing learnings or issue files listed above).
For each issue, invoke the report-issue skill:
Skill learning-agents:report-issue $ARGUMENTS "<one-sentence description>" "<timestamp of relevant line(s)>"
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.
## 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.")