Imported from xiangri678/email-intelligence-digest-skill (
AGENTS.md). Install upstream withnpx skills add xiangri678/email-intelligence-digest-skill. Copyright stays with the author.
AGENTS.md
This file records project-specific instructions for AI coding agents.
Project Rules
- Keep every published skill free of personal paths, account identifiers, database IDs, channel IDs, credentials, and private output.
- Preserve third-party attribution and licenses. Do not claim imported or bundled skills as original work.
- Keep
SKILL.mdfocused; move conditional procedures to a directly linked file underreferences/.
Architecture Notes
- Each skill is independently installable from its directory under
skills/. - Scripts must resolve sibling resources relative to their own file, never from a specific user's home directory.
Data and API Contracts
- External writes require user authorization and destination read-back.
- Public web acquisition must not read browser profiles or bypass access controls.
Validation Commands
- Run the skill-creator
quick_validate.pyagainst every changed skill. - Compile changed Python scripts with
python3 -m py_compile. - Run script help and non-mutating smoke checks where the host platform supports them.
Known Pitfalls
- Hermes CLI flags and scheduler behavior vary by release; verify them against the installed version.
- A successful job trigger or HTTP response is not proof that the intended work completed.
Debugging Playbook
- Inspect the smallest failing stage, then verify the corresponding local artifact and external read-back independently.
Do Not Change Without Care
- Do not add real credentials, captured cookies, personal database schemas, or private case data to examples or tests.
