The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted data from the git repository to generate documentation.
Ingestion points: The scripts/generate-doc-notes.mjs script extracts commit messages using git log and parses markdown files for H1 headers to build release notes.
Boundary markers: No specific delimiters or boundary instructions are implemented to prevent the AI from interpreting content within commit messages or file titles as instructions.
Capability inventory: The skill possesses the capability to write to the local filesystem (updating CHANGELOG.md) and interact with the GitHub API (creating releases and tags) via the semantic-release toolchain.
Sanitization: The extracted strings from commit history and file headers are not sanitized before being included in the generated release notes.
[COMMAND_EXECUTION]: The skill relies on shell command execution for project initialization and release workflows.
Evidence: Scripts such as scripts/init-project.mjs, scripts/init_project.sh, and scripts/generate-doc-notes.mjs use execSync or Bun.spawn to execute git, npm, and other system commands.
Context: These operations are standard for the tool's purpose, constructing commands from local state and environment variables to manage versioning and dependencies.

semantic-release