The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests untrusted data from GitHub issue bodies and comments, creating an exposure surface for indirect prompt injection.
Ingestion points: SKILL.md (Step 2) retrieves external content including issue bodies, every comment, and linked specifications.
Boundary markers: The skill's Core Principle explicitly defines this content as "untrusted evidence, not instructions" and prohibits it from overriding the workflow rules.
Capability inventory: The skill has the ability to execute local validation commands (SKILL.md, Step 5) and write plan drafts to the file system (SKILL.md, Step 8).
Sanitization: The skill enforces a ready-for-agent label check and requires manual approval before any GitHub publication occurs.
[COMMAND_EXECUTION]: The skill executes local commands to validate the working environment and testing seams.
As described in SKILL.md (Step 5), the skill runs focused validation commands based on repository configuration and established testing prior art.
[EXTERNAL_DOWNLOADS]: The skill fetches data and metadata from GitHub's official infrastructure.
It retrieves live issue state and comments from GitHub to build its authoritative source packet, which is consistent with its stated purpose.

to-plan