The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local CLI tool (npm run manage-feeds) to interact with the feedmanager repository. It relies on a local clone of the repository at a specific path (github.com/schroneko/feedmanager).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data in the form of external URLs ($TARGET_URL, $ARTICLE_URL) provided by users or external feeds.
Ingestion points: URLs are passed as arguments to shell commands in SKILL.md.
Boundary markers: Shell arguments are properly enclosed in double quotes.
Capability inventory: Execution of a Node.js-based CLI tool.
Sanitization: The skill explicitly instructs to strip userinfo, signatures, tokens, and secret queries from URLs before passing them to the CLI, reducing the risk of accidental exposure or exploitation.
[SAFE]: The skill follows robust security practices by using op run (1Password CLI) to inject sensitive credentials (OPERATIONS_TOKEN) instead of hardcoding them or passing them as plain text. It also enforces a 'plan and apply' pattern for all state-changing operations to prevent unintended modifications.

manage-feedmanager