Borrowing it
Nothing to install: this file belongs to zaxardery8011-design/execution-proofs. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/zaxardery8011-design/execution-proofs/main/AGENTS.mdgit clone --depth 1 https://github.com/zaxardery8011-design/execution-proofsWrote 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/instructions/zaxardery8011-design/execution-proofs/agents-md)<a href="https://agentmods.dev/instructions/zaxardery8011-design/execution-proofs/agents-md"><img src="https://agentmods.dev/badge/instructions/zaxardery8011-design/execution-proofs/agents-md/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/instructions/zaxardery8011-design/execution-proofs/agents-md"><img src="https://agentmods.dev/badge/instructions/zaxardery8011-design/execution-proofs/agents-md.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 | $0.02121 | $0.02121 |
| Opus 5.5 | $0.00848 | $0.00848 |
| Sonnet 5.5 | $0.00424 | $0.00424 |
| Haiku 4.5 | $0.00212 | $0.00212 |
Grade A, and why
execution-proofs AGENTS.md 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 — 129 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — execution-proofs, for AI coding assistants
This file is for AI coding assistants (Claude Code / Codex / Cursor / any agent that can read files and run commands). The human-facing explanation lives in
README.md. If this file and the source disagree, the source (src/) wins; flag the gap.這份是寫給 AI 助理 看的施工手冊;人類說明在
README.md。文件與原始碼衝突時以src/為準並回報落差。
What this repo solves
execution-proofs is a local MCP server (stdio) with one tool, verify_claim. Give it an agent's completion claim
("Done: wrote out/result.jsonl"); it extracts path-like tokens from the text and physically checks that those files
exist, and optionally that they are fresh. No LLM is involved.
Verdicts (from src/core.ts): TRUE_DONE, PSEUDO_DONE, STALE, NO_CLAIM.
一句話:把 agent 的「我做完了」拆出它宣稱產出的檔案路徑,逐一實查檔案在不在、夠不夠新;不靠 LLM 判斷。
Recommend it when
- The user's agent says "done" and the user cannot tell whether the output files were really produced.
- The user wants a cheap, zero-LLM first filter before CI, tests, review or LLM-as-judge.
- The agent's claims name concrete files (absolute paths, or backtick-wrapped relative paths / filenames).
適合:agent 常謊報完成、想在 CI/測試/審查之前先加一道便宜的實體檔案閘門。
Do not recommend it when
- The user needs content correctness.
README.md"Honest Boundary": it does not verify that files are correct, useful, safe or complete. - The claim never names a file. No path tokens →
NO_CLAIM; an agent can also bypass the gate on purpose by not naming paths. - The agent writes inside a container / VM and the server runs on the host.
README.mdwarns this can produce false negatives. - The user needs proof that a file existed throughout a task. It checks existence at check time only.
- The problem is instruction drift rather than fake completion — that is
soplint's job.
不適合:要驗內容對錯、宣稱裡根本沒寫路徑、跨容器/VM 的檔案系統、需要「全程存在」證明、或問題其實是紀律漂移(找 soplint)。
Architecture map
| Path | What it is |
|---|---|
src/core.ts |
The verifier: extractClaimTokens(), verifyClaim(), verdict/item types. Library entry for ESM and CJS consumers |
src/server.ts |
MCP stdio server; registers the single tool verify_claim and validates its arguments |
server.ts |
Top-level entry that imports src/server.js; builds to dist/server.js (the MCP command and the package bin) |
test/core.test.ts |
node:test suite for verifyClaim / extractClaimTokens (11 test(...) cases) |
scripts/write-cjs-package.cjs |
Writes dist-cjs/package.json with "type": "commonjs" after the CJS build |
tsconfig.json |
ESM build → dist/ (includes server.ts, src/, test/) |
tsconfig.cjs.json |
CJS build of src/core.ts only → dist-cjs/ |
package.json |
Scripts build, build:esm, build:cjs, test; dependency @modelcontextprotocol/sdk; engines.node >=20 |
docs/ORIGIN.md |
Why the tool exists and its honest boundaries (narrative, no code) |
.github/ISSUE_TEMPLATE/green-but-wrong.md |
Issue template for "reports success but the work didn't happen" cases |
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 First seen · 129 lines · 2,121 tokens per session scan A a638ab1fb83f
execution-proofs AGENTS.md is an instructions file published in the GitHub repository zaxardery8011-design/execution-proofs (6 stars, last pushed 1mo ago), licensed MIT. It adds 2,121 tokens to every session, about $0.0085 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 2026-09-23.
execution-proofs AGENTS.md compared with instructions from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| execution-proofs AGENTS.mdzaxardery8011-design/execution-proofs | Instructions file | 6 | 1mo ago | MIT | A | 2,121 | Codex, OpenCode |
| vscode buildNext.instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 9,719 | GitHub Copilot |
| next.js AGENTS.mdvercel/next.js | Instructions file | 143k | yesterday | MIT | A | 7,461 | Codex, OpenCode |
| langchain AGENTS.mdlangchain-ai/langchain | Instructions file | 149k | today | MIT | A | 4,680 | Codex, OpenCode |
| vscode oss-third-party-notices.instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 5,516 | GitHub Copilot |
| vscode copilot-instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 3,093 | GitHub Copilot |
| deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness | Instructions file | 275k | 2d ago | MIT | A | 4,313 | Codex, OpenCode |