The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local shell commands such as git diff, ls, and echo to inspect the local repository state and create fragment files. These operations are transparently documented and scoped to the project's directory structure.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the repository (e.g., filenames, diffs, and PR content) to generate changelog entries. It features a 'Mandatory Human Confirmation Gate' requiring explicit user approval before any write operation, which effectively mitigates the risk of the agent being misled by instructions embedded in untrusted data. Mandatory Evidence Chain: (1) Ingestion points: git diff and file listings in SKILL.md; (2) Boundary markers: None explicitly mentioned in templates; (3) Capability inventory: File write access to specific changelog directories and bash tool execution; (4) Sanitization: A mandatory human-in-the-loop confirmation gate requiring user approval for all actions.
[EXTERNAL_DOWNLOADS]: The skill references established industry standards such as keepachangelog.com and semver.org. These are well-known and trusted resources used for informational purposes.
[SAFE]: Overall, the skill demonstrates secure design patterns, including least privilege for tool usage and mandatory human-in-the-loop validation for all state-changing actions. No malicious patterns or security vulnerabilities were detected.

prowler-changelog