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 skills add sdfdu/evopilot --skill extension-foundrygit clone --depth 1 https://github.com/sdfdu/evopilotWrote 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/skills/sdfdu/evopilot/extension-foundry)<a href="https://agentmods.dev/skills/sdfdu/evopilot/extension-foundry"><img src="https://agentmods.dev/badge/skills/sdfdu/evopilot/extension-foundry/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/skills/sdfdu/evopilot/extension-foundry"><img src="https://agentmods.dev/badge/skills/sdfdu/evopilot/extension-foundry.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 | $0.00034 | $0.00473 |
| Opus 5.5 | $0.00014 | $0.00189 |
| Sonnet 5.5 | $0.00007 | $0.00095 |
| Haiku 4.5 | $0.00003 | $0.00047 |
Grade A, and why
extension-foundry 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
Extension Foundry
Evolve capability through promotion, testing, and rollback rather than uncontrolled self-modification.
- Run
evopilot_analyze_sequences; useevopilot_analyze_habitsonly to inspect individual actions. Ignore patterns supported by fewer than three observations. - At three observations, describe a candidate workflow. At five with at least three successful outcomes, it becomes
draft_ready. At eight with at least three successes, consider it stable. - Prefer a Skill when existing tools are sufficient. Add a deterministic script for repeatable mechanics.
- Draft an MCP server only when live data, authentication, controlled external actions, or a missing tool genuinely requires one.
- Use
evopilot_compile_skillonly for adraft_readyorstablefingerprint. Run bothevopilot_validate_skillandevopilot_assess_skill_quality, then refreshQUALITY_REPORT.mdwithevopilot_annotate_skill_qualityafter edits. ReviewSKILL.md,evopilot.json, and every quality annotation; keep the bundle uninstalled, version it, and test it in an isolated workspace. Structural validation is not proof that the workflow executes correctly. - Compare success rate, retries, corrections, time, and user acceptance against the previous workflow.
- Do not promote pure generic tool sequences or adjacent duplicate actions. When a meaningful workflow reaches
draft_readyorstable, proactively explain its evidence before generating it. Compile, validate, assess, and annotate locally, then present the exact installation approval only if both gates pass. After the user explicitly confirms, consume the one-time approval and install the Skill. Never install silently or treat an earlier general preference as installation approval. - Roll back when the new workflow performs worse or routes unrelated requests.
Never modify EvoPilot's approval policy through learned behavior. Explicit user instructions outrank learned memories.
When the PreToolUse gate blocks an exact action, show its approval ID and ask the user. Call evopilot_authorize_once only after the user explicitly confirms that action, then retry it once.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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 eedbb39adcae
- 1mo ago First seen · 22 lines · 34 tokens per session scan A e969a55ea0af
extension-foundry is a skill published in the GitHub repository sdfdu/evopilot (2 stars, last pushed 1mo ago), licensed MIT. It adds 34 tokens to every session and 473 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 skills, from other repositories
source-verify
A source-checking workflow for confirming that medical information comes from an authoritative, complete, and currently valid document.
cvm-ai-doctor
A health-diagnosis workflow for servers, computers, virtual machines, and containers on Linux, macOS, or Windows.
wechat-publisher-personal
A personal publishing skill that turns one Markdown article into posts for a Chinese content platform and RedNote, a Chinese social-media platform. It formats the long article for the first platform and adapts it into a shorter post with images and hashtags for RedNote.
clinical-q-and-a
A fast path for answering general educational medical questions and correcting common misconceptions, with a light check against one current authoritative guideline or consensus.
news
A news-finding guide that uses specified Chinese news websites for politics, business, society, international affairs, technology, sports, and entertainment.
guideline-learning-diagnosis
A study aid for learning a medical guideline or its diagnostic criteria. It helps assess what someone knows, find weak sections, and plan further study without diagnosing a real patient.