The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted content from external plan files, API responses, and third-party documentation.
Ingestion points: Processes content from .md plan files and external API sources during the freshness check and task execution phases (SKILL.md).
Boundary markers: Includes a dedicated 'untrusted-content-contract' section that mandates the use of <EXTERNAL_DATA> delimiters and explicit instructions to ignore instructions embedded within retrieved data.
Capability inventory: The skill utilizes shell-level tools including curl for network requests, git log for repository history, and grep for codebase inspection, alongside the ability to execute arbitrary tasks defined in a plan.
Sanitization: Explicitly forbids the agent from following requests in external data to run tools, reveal secrets, or change policy.
[COMMAND_EXECUTION]: The skill performs shell command execution for its 'Freshness check' logic, specifically using curl, git log, and grep to validate the authoritative state of the project before beginning implementation.

executing-plans