An agent skill for mapping undocumented website APIs with Chrome CDP, curl, and JS bundle analysis.
This repository contains the api-mapper skill package. The skill guides an agent through
capturing browser traffic, probing discovered endpoints, reverse-engineering client-side request
logic, and writing enough documentation to build a client without opening the browser again.
api-mapper/
SKILL.md
agents/
openai.yaml
references/
api-doc-requirements.md
cdp-capture.md
chrome-linux.md
chrome-macos.md
chrome-windows.md
flow-script-template.md
js-flow-analysis.md
- Python 3 + pip
- Google Chrome
- curl or
curl.exe
# Codex
npx -y skills add github.com/11philip22/api-mapper-skill -a codex --global --skill api-mapper
# opencode
npx -y skills add github.com/11philip22/api-mapper-skill -a opencode --global --skill api-mapper
# Claude
npx -y skills add github.com/11philip22/api-mapper-skill -a claude --global --skill api-mapperAfter the skill is installed, ask your agent to use it:
Use the api-mapper skill to reverse-engineer and document https://example.com.
The agent will open a real Chrome session, capture XHR/Fetch traffic, inspect JS bundles when requests contain computed values, and probe endpoints directly with curl.
Note
Chrome runs visibly. The user only needs to step in for credentials, CAPTCHAs, rate limits, or other walls the agent cannot pass.
The skill writes findings into the target project, not this skill repository:
artifacts/{timestamp}/
requests.ndjson
bundles/
docs/apis/
README.md
{area}.md
scripts/
flows/
API docs are language-agnostic. Complex browser-derived flows also get a runnable Python replay
script under scripts/flows/.