The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the gh (GitHub CLI) to perform authenticated operations like fetching repository information, release metadata, and PR details. This is the primary functionality and is executed within the scope of the user's current repository context.
[DATA_EXFILTRATION]: The skill fetches data from GitHub and displays it in the agent's chat context. While it uses the gh CLI to interact with external GitHub APIs, these operations are targeted towards the specific repository identified by the user or current context and do not show signs of malicious exfiltration.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection (Category 8).
Ingestion points: Untrusted data enters the agent context through the fetch-release-data.sh script, which retrieves title and body fields from GitHub Pull Requests.
Boundary markers: The SKILL.md instructions do not specify explicit delimiters or 'ignore' instructions for the agent when processing the fetched PR content.
Capability inventory: The agent has the capability to execute shell scripts (fetch-release-data.sh and validate-release-notes.sh) and use the gh CLI via the command line.
Sanitization: The scripts perform basic sanitization (stripping ANSI color codes using sed), but do not validate the semantic content of the PR titles or bodies to prevent instructions from influencing the agent's generation process.

write-release-notes