The Agent Skills Directory

[SAFE]: The skill is entirely instructional and focused on generating technical documentation. It provides templates and methodologies for writing migration guides without executing any external code or scripts.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes untrusted user data to generate documentation. However, the risk is mitigated by the skill's structure.
Ingestion points: User input gathered during the interview phase regarding project details and breaking changes (SKILL.md).
Boundary markers: The instructions require the use of specific templates (references/breaking-change-entry-template.md) and exact literal quoting of error messages, which helps prevent accidental interpretation of instructions.
Capability inventory: The skill writes output to migration guide documents and maintains a local memory file 'devrel-context.md'.
Sanitization: No specific sanitization logic is implemented beyond the instructional structure.
[SAFE]: Metadata and author information are consistent with the provided vendor context ('Samuel Berthe' / 'samber'). The skill references other related skills from the same author, which is a standard pattern for this developer's suite of tools.
[SAFE]: The skill mentions external tools and ecosystems (e.g., jscodeshift, cargo fix) only as examples or recommendations for users to include in their documentation, not for execution by the agent itself.

version-migration-guide