The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interacts with the local environment using several commands:
gh issue create and gh auth status are used to manage GitHub issues.
npm test is executed to run a local validation script (validate.mjs).
grep is used for de-duplication checks in local files.
A local script bin/wxsource.mjs is invoked to fetch content from external URLs (WeChat and Kanxue articles).
[DATA_EXFILTRATION]: The skill is capable of sending distilled experience data to an external GitHub repository (dslsdzc/rev-skills). While the workflow requires the gh CLI to be authenticated and includes a mandatory "Privacy Desensitization" step to remove sensitive project or company details, it establishes a functional path for data to leave the local environment.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources that could contain malicious instructions designed to influence the agent's behavior during the distillation process.
Ingestion points: External articles fetched via URL (WeChat, Kanxue), user notes, and session history.
Boundary markers: The skill enforces a specific distillation template and a "Privacy Desensitization" checklist to structure the output and ignore identifying details.
Capability inventory: The skill can modify local Markdown files (SKILL.md, references/experience.md), execute local commands (npm test, grep), and perform network operations via the GitHub CLI.
Sanitization: Mandatory rules in references/intake.md require stripping all specific names (company, product, internal codes) and泛化 (generalizing) the technical content before it is stored or shared.

re-feedback