The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a natural attack surface for indirect prompt injection as it processes user-provided content (Title, Context) and interpolates it into persistent project files.
Ingestion points: SKILL.md Step 1 gathers 'Category', 'Title', and 'Context' from the user.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are included in the ADR template.
Capability inventory: The skill writes new files to docs/adrs/ and modifies metadata files in .claude/data/adrs/.
Sanitization: No sanitization is performed on the descriptive context provided by the user before it is written to the project files.
This finding is categorized as safe because the ingestion and storage of user-provided text is the primary and intended purpose of a documentation skill.
[SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were detected. The skill operates exclusively on local project files and metadata to manage technical documentation as requested by the user.

secondbrain-adr