The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches and processes untrusted data from GitHub Pull Request titles and descriptions, which could be used to manipulate the agent's output during the release note drafting process.
Ingestion points: Step 4 utilizes gh pr list --json number,title,body to retrieve external content from merged Pull Requests.
Boundary markers: The skill lacks explicit instructions or delimiters to isolate the PR content or warn the agent to ignore potentially malicious instructions embedded within that data.
Capability inventory: The skill has significant local and remote capabilities, including executing project-specific checks via make check, committing/tagging the repository, and updating live GitHub releases via gh release edit.
Sanitization: There is no evidence of filtering or sanitizing the PR metadata before it is interpolated into the agent's context for summarization.
[COMMAND_EXECUTION]: The skill executes various shell commands to manage the software lifecycle.
Evidence: Uses git for repository state checks, tagging, and pushing changes.
Evidence: Uses gh (GitHub CLI) for querying Pull Requests and managing releases.
Evidence: Executes make check, which runs the local project's validation suite. This is a standard and expected part of a release workflow.

releasing