The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data by reading Pull Request descriptions, commit messages, and changed files from a product repository to generate changelogs and announcement drafts.
Ingestion points: references/shipped-changes.md instructs the agent to read PR titles/descriptions and git commit logs to identify what features have shipped.
Boundary markers: No specific delimiters or boundary markers are defined for the repository data ingestion to distinguish between instructions and data.
Capability inventory: The skill requires the capability to read files and git history via terminal tools and to generate marketing text drafts.
Sanitization: The skill mitigates risks by mandating that all generated drafts must be staged for human review before publishing and by instructing the agent to filter out invisible code changes.
[COMMAND_EXECUTION]: The skill utilizes git commands (git log) to analyze local repository history as described in references/shipped-changes.md. These commands are localized to the agent's workspace and are used for legitimate analytical purposes to determine product updates.

launch