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/rasputinkaiser/self-improvement-plugin/retrogit clone --depth 1 https://github.com/RasputinKaiser/Self-Improvement-PluginWhat 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 | $0.00022 | $0.00995 |
| Opus 5 | $0.00011 | $0.00498 |
| Sonnet 5 | $0.00004 | $0.00199 |
| Haiku 4.5 | $0.00002 | $0.00100 |
Grade A, and why
retro 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.
How it starts
The opening of the file, as written. The whole thing — 99 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/retro — close the loop on this session
Argument: $ARGUMENTS is an optional focus. Use it to narrow the review to a
repo, task, failure, or capability. A retrospective is a reusable learning
pass, not a diary entry and not a reason to manufacture a lesson.
1. Establish the evidence boundary
Resolve the current scope and collect the smallest useful evidence set:
- Run
pwd -P,git status --short, andgit diff --statfor the current repository. Treat existing dirty changes as user-owned unless the current task clearly explains them. - Query recent structured outcomes:
python3 ${PLUGIN_ROOT:-${CLAUDE_PLUGIN_ROOT}}/scripts/task_outcome_tracker.py --query --limit 8. - Search scoped prior lessons:
python3 ${PLUGIN_ROOT:-${CLAUDE_PLUGIN_ROOT}}/scripts/recall_ranker.py --query "$ARGUMENTS" --json. - Re-read the current conversation for the actual failed attempt, working correction, user direction, and final proof. If the host exposes an exact transcript path, preserve it as the evidence path; do not invent a path from a session id. If a separately installed Retro transcript miner is available, it may supplement this pass, but its absence is not a SIPS failure and this command must not duplicate that miner.
- Inspect
state.yamlor relevant receipts only when they can confirm or disprove a conclusion. Keep source, cache, runtime, host, and public proof layers separate.
2. Classify the signal
Rank fix-ups and recurring hurdles before new observations. For each candidate, identify the exact evidence and classify it as one of:
- record — a durable retry chain, unresolved failure, user correction, prerequisite, environment fact, or repeated friction that another session could reuse;
- skip — one-off noise, a typo, a transient provider/network issue, or a conclusion supported only by absence;
- unclear — plausible but not yet supported; name the missing proof.
An unresolved failure is not a success lesson. A workaround is not a fix unless the working form and its verification are both visible. Preserve the user's standing preferences separately from inferred technical facts.
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 · 99 lines · 22 tokens per session scan A b05c9784389d
retro is a command published in the GitHub repository RasputinKaiser/Self-Improvement-Plugin (6 stars, last pushed 5d ago), licensed MIT. It adds 22 tokens to every session and 995 once invoked, about $0.0001 per session on Opus 5. 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
quiz
Generate and play an interactive study quiz from a KMS knowledge base, URL, file, or topic — and record the score back into the KMS.
quiz-result
Analyze quiz progress over time and the weak areas to study next, from a KMS scores log.
retro
Write this session's lessons into the project's lessons/ directory using the memory-discipline format.
cast
Choose, build, or synthesize the right harness for the current task.
learn
Force claude-smart to extract learnings from this session now.
component
Scaffold a new React component grounded in the paper-mono primitives. Requires explicit kind or a nearest-existing-component match. No empty divs, no speculative scaffolding.