The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a bundled maintenance script scripts/check-digests.mjs using Node.js to verify the integrity of local skill files against recorded SHA256 hashes. This is a local, functional requirement for the skill's purpose.
[EXTERNAL_DOWNLOADS]: To check for updates, the skill connects to remote repository APIs or uses Git to fetch the latest revisions of third-party skills. The instructions specify that this must be done via official repository APIs or read-only Git inspection.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted content from external repositories which could contain malicious instructions designed to influence the agent.
Ingestion points: Fetches upstream skill files (instructions, scripts, metadata) from remote repositories; reads local skill-sources.json and 02-SKILLS.md configuration files.
Boundary markers: The skill contains strong boundary instructions, explicitly telling the agent to treat fetched instructions as comparison data only ("do not execute them... to check for updates") and defines an "Authority boundary" to prevent accidental execution of embedded instructions.
Capability inventory: The skill can execute the bundled Node.js script, perform network requests to fetch updates, and write or modify local files when in Update mode.
Sanitization: The workflow mandates manual review of all changed scripts and validation of changed instructions using provided evaluation guidance (references/evaluation.md) before any application or execution.

update-skills