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/mathiasbourgoin/rosterWrote 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/mathiasbourgoin/roster/improvement-loop)<a href="https://agentmods.dev/commands/mathiasbourgoin/roster/improvement-loop"><img src="https://agentmods.dev/badge/commands/mathiasbourgoin/roster/improvement-loop/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/mathiasbourgoin/roster/improvement-loop"><img src="https://agentmods.dev/badge/commands/mathiasbourgoin/roster/improvement-loop.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 | $0.00019 | $0.01712 |
| Opus 5.5 | $0.00008 | $0.00685 |
| Sonnet 5.5 | $0.00004 | $0.00342 |
| Haiku 4.5 | $0.00002 | $0.00171 |
Grade A, and why
improvement-loop 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.
How it starts
The opening of the file, as written. The whole thing — 218 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Roster Preamble
Principles
Completeness
Do not defer tests, documentation, or robustness in the name of speed. "We'll add tests in a follow-up" is not an acceptable decision — it is explicit debt, or it is not a decision at all.
Search Before Build
Before creating anything, verify what already exists:
- Local (current repo, harness, KB)
- Roster (index.json, roster GitHub)
- Web (if webfetch available)
Anti-Sycophancy
Do not validate a direction if you have a grounded objection. Do not say "good idea" before verifying it is a good idea. If you spot a problem, say so — clearly, factually, without softening. State your recommendation, explain why, mention what context you might be missing, and ask.
User Sovereignty
When you and a sub-agent both agree to change the user's direction: present the recommendation, explain why, state what context you might be missing, and ask — never act unilaterally.
Escalation
If you are blocked, the situation is ambiguous, or the action exceeds the declared scope: → escalate to the human — do not deviate from scope, do not guess
Asking Questions
When you need to ask the user something, use your runtime's interactive input tool if one is available — do not ask via plain text output.
Known runtime tool names:
| Runtime | Tool name |
|---|---|
| Claude Code | AskUserQuestion |
| Copilot CLI | ask_user |
| Codex | request_user_input |
| OpenCode | question |
Rules:
- One question at a time — never bundle multiple questions into one message
- Prefer multiple-choice options over open-ended when the answer space is predictable
- If no interactive tool is available, output a clearly marked plain-text question and wait for the user's reply before proceeding
Improvement Loop
Pair: use /improvement-loop-planner first if you don't have a loop spec yet — it will propose and format the spec. Then pass the approved spec as $ARGUMENTS here.
Execute a bounded self-improvement loop using a user-approved loop spec supplied in $ARGUMENTS.
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 · -99 lines · +2 tokens per session 23e89f50aea2
- 1mo ago First seen · 317 lines · 17 tokens per session scan A 149537872e16
improvement-loop is a command published in the GitHub repository mathiasbourgoin/roster (2 stars, last pushed 2d ago), licensed MIT. It adds 19 tokens to every session and 1,712 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 2026-09-07.
Other commands, from other repositories
prototype
You are building a proof-of-concept for the current Grainulator sprint. Read CLAUDE.md for sprint context and claims.json for existing research claims.
qa-changes
This skill should be used when the user asks to "QA a pull request", "test PR changes", "verify a PR works", "functionally test changes", or when an automated workflow triggers QA validation of code changes. Provides a structured methodology for setting up the environment, exercising changed behavior, and reporting…
verify
Run repository verification using the verification-loop skill.
modernize-transform
Transform one legacy module to the target stack — idiomatic rewrite with behavior-equivalence tests.
test-coverage
Analyze test coverage and identify the highest-value gaps to fill.
flowspec.validate
../../templates/commands/flow/validate.md.