The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses multiple shell commands to manage repository state, including git merge, git push, git branch -D (destructive deletion), and gh pr create. These commands are necessary for the skill's purpose but require user trust in the agent's actions.
[INDIRECT_PROMPT_INJECTION]: The skill has a potential attack surface for indirect prompt injection because it processes untrusted data from the repository's history and diffs.
Ingestion points: The skill reads output from git log and git diff during safety checks and when generating pull request bodies. It also reads configuration from 02-DOCS/wiki/sdd/config.yaml and user profiles.
Boundary markers: There are no explicit delimiters or instructions to help the agent distinguish between data and instructions when processing git history.
Capability inventory: The agent can perform significant repository changes, including deleting local and remote branches and pushing new commits.
Sanitization: The skill does not appear to sanitize data extracted from the repository before using it to construct commit messages or pull request descriptions.
[EXTERNAL_DOWNLOADS]: The skill references the execution of tools via npx, which involves downloading and running packages from the NPM registry.
Evidence: SKILL.md mentions npx @ericrisco/rsc sello off and npx @ericrisco/rsc worktrees.
Note: These packages are scoped to the skill's author ('ericrisco') and represent expected vendor-provided functionality.

ship