Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Dispatch entry point for the LearningAgents plugin. Routes to sub-commands for creating agents, running learning cycles, and reporting issues.
.claude/skills/majiayu000-learning-agents/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 803% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 441% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 263% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 44% | 0% |
Manage auto-improving AI sub-agents that learn from their mistakes across sessions.
$ARGUMENTS is the text after /learning-agents (e.g., for /learning-agents create foo, $ARGUMENTS is create foo).
Before routing, check if .claude/session_log_folder_info.md exists. If it does not exist, run Skill learning-agents:setup first, then continue with routing below.
Only perform this check once per session — after the setup skill completes (or if the file already exists), proceed directly to routing for all subsequent invocations.
Split $ARGUMENTS on the first whitespace. The first token is the sub-command (case-insensitive); the remainder is passed to the sub-skill. Accept both underscores and dashes in sub-command names (e.g., report_issue and report-issue are equivalent).
create <name>Create a new LearningAgent scaffold.
Invoke: Skill learning-agents:create-agent <name>
Example: $ARGUMENTS = "create rails-activejob" → Skill learning-agents:create-agent rails-activejob
learnRun the learning cycle on all pending session transcripts. Any arguments after learn are ignored.
Invoke: Skill learning-agents:learn
report_issue <agentId> <details>Report an issue with a LearningAgent from the current session.
Invoke: Skill learning-agents:report-issue <session_log_folder> <details>
To construct the session log folder path: search .deepwork/tmp/agent_sessions/ for a subdirectory whose name contains the provided agentId. The path structure is .deepwork/tmp/agent_sessions/<session_id>/<agentId>/. If no match is found, inform the user. If multiple matches exist, use the most recently modified one.
Example: $ARGUMENTS = "report_issue abc123 Used wrong retry strategy" → find folder matching abc123 under .deepwork/tmp/agent_sessions/, then Skill learning-agents:report-issue .deepwork/tmp/agent_sessions/sess-xyz/abc123/ Used wrong retry strategy
Display available sub-commands:
LearningAgents - Auto-improving AI sub-agents
Available commands:
/learning-agents create <name> Create a new LearningAgent
/learning-agents learn Run learning cycle on pending sessions
/learning-agents report_issue <agentId> <details> Report an issue with an agent
Examples:
/learning-agents create rails-activejob
/learning-agents learn
/learning-agents report_issue abc123 "Used wrong retry strategy for background jobs"$ARGUMENTS doesn't match any known sub-command, show the help text above| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→fail | 3,466 | 7,142 | +106% | 1 | 1 | 0% | 511 | 1,441 | +182% | 0 | 0 | — |
case-08 | fail→fail | 18,964 | 20,541 | +8% | 1 | 1 | 0% | 3,542 | 2,583 | -27% | 0 | 0 | — |
case-01 | fail→pass | 18,126 | 14,680 | -19% | 1 | 1 | 0% | 2,132 | 2,492 | +17% | 0 | 0 | — |
case-02 | fail→fail | 4,665 | 13,312 | +185% | 1 | 1 | 0% | 680 | 1,911 | +181% | 0 | 0 | — |
case-04 | fail→fail | 4,045 | 10,642 | +163% | 1 | 1 | 0% | 607 | 1,533 | +153% | 0 | 0 | — |
case-05 | fail→fail | 7,674 | 15,529 | +102% | 1 | 1 | 0% | 1,379 | 2,578 | +87% | 0 | 0 | — |
case-06 | fail→pass | 7,570 | 21,536 | +184% | 1 | 1 | 0% | 369 | 3,332 | +803% | 0 | 0 | — |
case-07 | fail→fail | 7,641 | 36,359 | +376% | 1 | 1 | 0% | 421 | 7,326 | +1640% | 0 | 0 | — |
case-09 | fail→pass | 5,004 | 22,496 | +350% | 1 | 1 | 0% | 774 | 4,188 | +441% | 0 | 0 | — |
case-10 | fail→pass | 7,631 | 23,004 | +201% | 1 | 1 | 0% | 1,149 | 4,173 | +263% | 0 | 0 | — |
case-11 | fail→pass | 16,325 | 8,866 | -46% | 1 | 1 | 0% | 1,713 | 2,464 | +44% | 0 | 0 | — |
case-12 | fail→pass | 3,125 | 7,165 | +129% | 1 | 1 | 0% | 313 | 1,063 | +240% | 0 | 0 | — |
case-13 | pass→pass | 12,949 | 3,696 | -71% | 1 | 1 | 0% | 1,246 | 1,295 | +4% | 0 | 0 | — |
case-14 | pass→pass | 21,984 | 26,565 | +21% | 1 | 1 | 0% | 3,609 | 3,015 | -16% | 0 | 0 | — |
case-15 | pass→fail | 5,076 | 22,117 | +336% | 1 | 1 | 0% | 731 | 1,619 | +121% | 0 | 0 | — |
case-16 | pass→fail | 11,618 | 5,920 | -49% | 1 | 1 | 0% | 2,296 | 1,839 | -20% | 0 | 0 | — |
case-17 | pass→fail | 5,576 | 5,715 | +2% | 1 | 1 | 0% | 951 | 1,076 | +13% | 0 | 0 | — |
case-18 | fail→pass | 10,443 | 6,462 | -38% | 1 | 1 | 0% | 2,020 | 1,949 | -4% | 0 | 0 | — |
case-19 | fail→fail | 8,820 | 19,086 | +116% | 1 | 1 | 0% | 591 | 1,535 | +160% | 0 | 0 | — |
case-20 | pass→pass | 18,415 | 16,791 | -9% | 1 | 1 | 0% | 2,592 | 3,006 | +16% | 0 | 0 | — |
case-21 | pass→fail | 18,605 | 14,991 | -19% | 1 | 1 | 0% | 2,547 | 3,555 | +40% | 0 | 0 | — |
case-22 | pass→pass | 16,442 | 13,260 | -19% | 1 | 1 | 0% | 2,221 | 2,337 | +5% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 15 counted toward the lift figure. The other 7 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +14 percentage points is the difference between those two pass rates over the 15 comparable cases. 4 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.