The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests commit messages from external repositories to classify release types. This creates a surface where malicious instructions embedded in a commit message could theoretically influence the agent's logic during the classification or recommendation phase.
Ingestion points: Commit messages are fetched via gh api in Step 2 of the workflow.
Boundary markers: The skill does not define specific delimiters to isolate commit messages from instructions, though the workflow includes a mandatory human-in-the-loop confirmation step in Step 4.
Capability inventory: The skill has the capability to create GitHub releases (gh release create) and trigger remote workflows (gh workflow run).
Sanitization: No explicit sanitization or filtering of commit content is described in the logic.
[COMMAND_EXECUTION]: The skill performs multiple operations via the gh command-line interface, including repository interaction and workflow triggering.
Evidence: Executes gh release create to tag repositories and gh workflow run to trigger a remote Homebrew tap update.
[EXTERNAL_DOWNLOADS]: The skill specifies a dependency on the gh oss-watch extension, which is an external tool not bundled with the skill itself.
Evidence: The compatibility field and Step 1 reference the requirement for the gh oss-watch extension.

gh-oss-release