The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from git commit history to generate user-facing release notes, creating an indirect prompt injection surface. Evidence: (1) Ingestion points: Git commit messages are collected via git log (SKILL.md). (2) Boundary markers: The skill instructions require showing a full preview and asking for user confirmation before writing to CHANGELOG.md (SKILL.md, workflow step 9). (3) Capability inventory: The skill can read git history and write to the filesystem. (4) Sanitization: Instructions guide the agent to translate technical messages into user-facing prose and filter out housekeeping commits, which acts as a sanitization layer.
[COMMAND_EXECUTION]: The skill uses shell commands (git log) with user-provided arguments for the range boundary. The instructions to 'use it directly' (SKILL.md, mode 1, step 2) represent a potential command injection surface if the underlying agent platform does not sanitize shell arguments.
[EXTERNAL_DOWNLOADS]: The README provides installation instructions using npx to fetch the skill from the npm registry, which is a well-known service.

changelog