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.
git clone --depth 1 https://github.com/RasputinKaiser/Self-Improvement-PluginWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/rasputinkaiser/self-improvement-plugin/selfloop)<a href="https://agentmods.dev/commands/rasputinkaiser/self-improvement-plugin/selfloop"><img src="https://agentmods.dev/badge/commands/rasputinkaiser/self-improvement-plugin/selfloop/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/rasputinkaiser/self-improvement-plugin/selfloop"><img src="https://agentmods.dev/badge/commands/rasputinkaiser/self-improvement-plugin/selfloop.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00013 | $0.00334 |
| Opus 5.5 | $0.00005 | $0.00134 |
| Sonnet 5.5 | $0.00003 | $0.00067 |
| Haiku 4.5 | $0.00001 | $0.00033 |
Grade A, and why
selfloop 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 1mo 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
Follow references/command-protocol.md.
Use goal_state.py selfloop-set <focus> only for an explicitly requested loop.
status, pause, resume, complete, and clear retain their documented state
controls. A request for status never starts a cycle. No background scheduler or
automatic child-agent campaign is implied.
For each authorized foreground cycle: inspect self_correct.py --json and
reproducible failures; use /checkpoint for recoverable state; create an adaptation
episode; declare competing hypotheses; inspect diagnostic_design for hypotheses
that no proposed probe can separate; gather evidence or propose one intervention.
The active agent authors the isolated candidate and evaluates it against the
frozen original, consumer, and counterexample checks.
Record review-ready, rejected, unavailable, or plateau evidence. Use
selfloop-record only with the corresponding supported result; evaluation does
not mean an improvement is deployed. Stop at ready_for_review for generated
changes. Use /teach for a durable candidate lesson, not every iteration.
A plateau means no supported next intervention within the current evidence and budget; it is not proof that no improvement exists. Keep one candidate at a time and the episode's persisted limits. Honor pause/stop immediately and report an external blocker accurately without silently changing user goal controls.
Use adaptation view next and action advance for the resumable foreground
cycle. Match packet identity and required output exactly. Link later evaluated
uses with link_outcome; report unmatched comparisons as unmeasured.
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.
- 1mo ago Changed · -44 lines · -16 tokens per session b34cdc81d5b4
- 1mo ago First seen · 75 lines · 29 tokens per session scan A abd41bbf619a
selfloop is a command published in the GitHub repository RasputinKaiser/Self-Improvement-Plugin (8 stars, last pushed 16d ago), licensed MIT. It adds 13 tokens to every session and 334 once invoked, about $0.0001 per session on Opus 5.5. 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
feature
Orchestrate a complete feature through discovery, spec, implementation, and review.
test
Design or run focused test validation for a task, bug, or diff.
research
Research a technical or product question.
prp-plan
Create comprehensive feature implementation plan with codebase analysis and pattern extraction.
santa-loop
Adversarial dual-review convergence loop — two independent model reviewers must both approve before code ships.
gan-design
Run a generator/evaluator design loop for frontend or visual work with bounded iterations and scoring.