The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using git and gh (GitHub CLI) to retrieve commit logs, view pull request details, and manage code versioning (staging, committing, and pushing). These operations target the vendor's own infrastructure (forcedotcom).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text from pull request titles and bodies that could contain malicious instructions. 1. Ingestion points: PR metadata is dynamically retrieved via gh pr view and read from the local CHANGELOG.md file. 2. Boundary markers: No specific delimiters or safety instructions are defined to separate the untrusted PR content from the agent's operational instructions. 3. Capability inventory: The agent has the ability to modify files within the repository and push those modifications to the remote server via git push. 4. Sanitization: The risk is mitigated by the instruction to rewrite and polish the content into a specific user-facing format and the mandatory requirement for explicit human approval before any changes are committed or pushed.

changelog