Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/trtmn/agent-plugins/self-improvementgit clone --depth 1 https://github.com/trtmn/agent-pluginsWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.00427 |
| Opus 5 | $0.00000 | $0.00214 |
| Sonnet 5 | $0.00000 | $0.00085 |
| Haiku 4.5 | $0.00000 | $0.00043 |
Grade A, and why
self-improvement scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
Run the self-improvement review pipeline now, in the foreground, against this machine's pending learnings.
First, check that ~/.claude/agents/learnings.md and ~/.claude/agents/learning-investigator.md exist. If either is missing, stop and tell me: "⚠ The self-improvement agents aren't wired up. Run /self-improvement:setup first." Do not proceed without them.
Then delegate to the self-improvement subagent via the Agent tool (subagent_type: self-improvement). Run it in the foreground so I can watch promotions happen live and interrupt if needed. Do NOT pass a transcript path (this is a mid-session manual run, not a session-end sweep) unless I explicitly ask you to sweep this conversation.
The subagent runs the same non-interactive pipeline the autonomous session-end hook uses:
- (Optional, only if I ask) sweep this conversation for anything capture missed.
- Read pending entries from
~/.learnings/{LEARNINGS,ERRORS,FEATURE_REQUESTS}.md— and, since this is a manual run inside a repo, the project-level.learnings/mirror too. - Dispatch a
learning-investigatorper candidate to judge needed-ness (conservative bar: high confidence + a recurrence signal + not a duplicate). - Auto-promote the qualifiers — append to the target CLAUDE.md, log a
[PROMO-<hex>]to~/.learnings/CHANGELOG.md, remove from the pending file. Skip clear rejects with a[SKIP-<hex>]. Leave uncertain entries pending. - Report a summary: swept / promoted / skipped / left-pending, with targets.
This run auto-promotes (no per-item approval prompt) — that's intentional and matches the autonomous behavior. Everything is logged; undo any promotion with /self-improvement:revert <PROMO-hex>.
For raw autonomous capture (the passive side), use the separate learnings subagent — not this one.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 17 lines · 0 tokens per session scan A d42ca33fbdc3
self-improvement is a command published in the GitHub repository trtmn/agent-plugins (2 stars, last pushed 5d ago), licensed Unlicense. It costs nothing until one of its globs matches a file; then it loads 427 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
claude-flow-swarm
Coordinate multi-agent swarms for complex tasks.
advisor
Advisory gate for triage or plan decisions. Spawns a second-opinion agent that challenges assumptions, surfaces risks, and proposes alternatives before the decision commits. Based on Anthropic advisor tool pattern.
bootstrap
PACT session-start ritual — identify the session team (platform-provisioned), secretary spawn, paused-state surface, bootstrap marker.
ox-session-review
Command "ox-session-review" from sageox/ox, covering failure-mode watch-list (read first), from the ledger root. should print 0, phase 1 — scan & score (read-only), quality buckets (first match wins) and removal candidates.
add-agent
引导新增一个 Agent 适配器。用法 /add-agent.
migrate
Command "migrate" from chohra-med/expo_boilerplate, covering invocation, steps and rule.