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 instructions/mperkhou/career-agent-workbench/agents-mdgit clone --depth 1 https://github.com/mperkhou/career-agent-workbenchWrote 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/mperkhou/career-agent-workbench/agents-md)<a href="https://agentmods.dev/instructions/mperkhou/career-agent-workbench/agents-md"><img src="https://agentmods.dev/badge/instructions/mperkhou/career-agent-workbench/agents-md.svg" alt="Measured on agentmods" 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 | $0.00452 | $0.00452 |
| Opus 5 | $0.00226 | $0.00226 |
| Sonnet 5 | $0.00090 | $0.00090 |
| Haiku 4.5 | $0.00045 | $0.00045 |
Grade A, and why
career-agent-workbench 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 4d 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 — 42 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent Guidance
Use a Python-first architecture and public-safe synthetic evidence in this repository.
- Never add real profile or application state, generated career artifacts, credentials, employer-internal material, or machine-specific paths.
- Keep private workspace state in the root-anchored ignored directories and
preserve the root
.envprotections. - Keep edits scoped, add focused tests, and align the version, changelog, and release note for committed changes.
- Do not use destructive Git operations or discard work without explicit authority.
- Preserve human-review checkpoints and require explicit authority before actions that publish, contact third parties, or change external state.
Local operator contract
- Run
make installfrom the repository root to create the ignored public.venv, install.[dev,browser], and install Playwright Chromium. Do not use the legacy repository environment for public operation. - The ignored root
.envis bootstrap-only. It selects the external workspace and its private dotenv; tokens and machine-specific settings belong only in that external mode-0600file. - Use
make start-website,make stop-website, andmake restart-websitefor PID-owned local lifecycle. Browser opening is opt-in throughOPEN_BROWSER=1; never stop processes by port scanning. make skill-linkmanages exactly the five public repository skills. It must not replace unrelated personal skills or overwrite a non-symlink.- The canonical MCP executable is the absolute public
.venv/bin/career-agent-workbench-mcp; retain a clearly named legacy registration until the final cutover gate.
Validation and release
- Focused changes require their tests, Ruff check/format,
git diff --check, and public-tree safety. Release readiness additionally requires the complete suite, wheel/sdist build and scan, isolated wheel smoke, offline Chromium, browser workbench, and injected MCP stdio checks. - Release PRs stay unmerged and untagged until their final gate and explicit user approval. Preserve normal reviewed commits; do not squash or rebase them.
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.
- 4d ago First seen · 42 lines · 452 tokens per session scan A b24a2adcefa2
career-agent-workbench AGENTS.md is an instructions file published in the GitHub repository mperkhou/career-agent-workbench (0 stars, last pushed 1mo ago), licensed MIT. It adds 452 tokens to every session, about $0.0023 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 instructions, from other repositories
gitprofile AGENTS.md
AGENTS.md instructions for arifszn/gitprofile, covering agents.md, what this project is, customizing a portfolio: edit one file, the base option is the most common source of breakage and layout.
personal-site AGENTS.md
Instructions for mldangelo/personal-site, covering agents.md, quick commands, project structure, code style and git workflow.
readykit AGENTS.md
AGENTS.md instructions for level09/readykit, covering agents.md, stack, commands, database migrations and review migrations/versions/ .py, adjust.
portfolio-nextjs AGENTS.md
Instructions for nixrajput/portfolio-nextjs, covering ai agent guidelines, project, layout, things that will bite you and adding a brand palette.
openalgo CLAUDE.md
Claude Code instructions for marketcalls/openalgo, covering claude.md, documentation map, skills, security and deployment model and runtime constraints.
bloomfolio CLAUDE.md
Instructions for lauroguedes/bloomfolio, covering claude.md, project overview, development commands, install dependencies and start development server (localhost:4321).