The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content such as issue reports and pull request diffs. This creates a potential surface where malicious instructions embedded in a report or code could attempt to influence the agent's analysis or output.
Ingestion points: The agent is instructed to read reporter logs and PR content (as described in SKILL.md and README.md).
Boundary markers: No explicit instructions are provided to the agent to treat external text as untrusted or to ignore embedded instructions, though the workflow includes human review as a safeguard.
Capability inventory: The skill uses gh api and shell commands to interact with GitHub repositories, providing a path for actions based on interpreted input.
Sanitization: There is no explicit requirement for the agent to sanitize or escape data before processing or including it in drafted replies.
[COMMAND_EXECUTION]: The skill utilizes shell commands and the GitHub CLI (gh) to automate tasks like fetching pull request data, managing review cycles, and interacting with the GitHub API. These operations are performed within the user's shell environment to manage repository resources.
[DYNAMIC_EXECUTION]: To handle GitHub reviews, the skill generates temporary JSON files and uses utility commands like sed to modify payloads before submission. This runtime generation and manipulation of files is used to correctly format API requests.

reply-and-review-discipline