The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches and processes external content from GitHub PR comments and review threads, which creates a potential surface for indirect prompt injection where instructions embedded in findings could influence agent behavior.
Ingestion points: The agent retrieves thread comments and PR-level suggestions via the GitHub API (gh api) as defined in references/github-queries.md (§2, §4, and §8).
Boundary markers: There are no instructions to use delimiters or explicit "ignore embedded instructions" warnings when processing the retrieved external content.
Capability inventory: The skill allows the agent to execute git commit, git push, and gh api (to post replies and resolve threads), as well as running repository-specific tests, linting, and type-checking tools.
Sanitization: No sanitization or filtering of retrieved comment content is performed before the agent uses it for triage and fixing code.
[COMMAND_EXECUTION]: The skill requires the agent to execute shell commands to interact with the environment and GitHub API.
Instructions in SKILL.md and references/github-queries.md direct the agent to use git (push, add, commit) and the GitHub CLI (gh) for authentication, viewing PRs, and performing complex GraphQL and REST API queries.
[DYNAMIC_EXECUTION]: The skill includes a test suite that executes system commands at runtime to verify its logic.
The file tests/query_tests.py uses the Python subprocess.run function to execute jq and bash to verify the documented API query filters.
A tests/run.sh script is provided to facilitate execution of this test suite.

kodus-loop