The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted repository data, including issue templates, labels, and maintainer names, retrieved via the kscripts gh-orient tool. This data is used to populate GitHub issue and pull request bodies.
Ingestion points: SKILL.md (Step 2: Repository Orientation via kscripts gh-orient).
Boundary markers: Absent for external repository data; the skill relies on instructional constraints (e.g., "Use orientation output strictly for taxonomy") and a mandatory human approval gate.
Capability inventory: gh issue create, gh pr create (Step 5: Execution).
Sanitization: No explicit sanitization or escaping of repository-provided strings is documented.
[COMMAND_EXECUTION]: The skill utilizes shell commands to orient to repositories and perform submissions.
Evidence: The workflow executes kscripts gh-orient, kscripts pr-check, tail, and the gh (GitHub CLI) tool. These operations are gated by a mandatory user approval step (ask_question in Step 4) before any remote execution occurs.

gh-post