The Agent Skills Directory

[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by processing external data from GitHub release notes.
Ingestion points: The scripts/generate_changelog.js script reads release data from stdin during the changelog generation step.
Boundary markers: No delimiters or ignore-instructions warnings are used when writing external content to the CHANGELOG.md file.
Capability inventory: The skill has the capability to perform high-impact actions including npm publish, git push, and gh release create (documented in SKILL.md).
Sanitization: The scripts/generate_changelog.js script does not perform any sanitization or validation of the r.body field before outputting it.
[COMMAND_EXECUTION]: The skill performs several powerful shell operations as part of its standard workflow.
The workflow executes npm publish, which can modify the state of public package registries.
The workflow executes git push and gh release create, which can modify the state of the remote repository and its releases.
The skill accesses and executes code from a non-standard local path (~/Scripts/claude-mem/), which may lead to security or stability issues if the environment contains unexpected or malicious scripts.

claude-code-plugin-release