The Agent Skills Directory

[SAFE]: The skill instructions describe a benign process for creating Architecture Decision Records. It requires explicit user approval before performing any write operations to the filesystem.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes existing files from the repository which could contain malicious instructions designed to influence the agent.
Ingestion points: The agent reads existing ADR files from repository paths such as docs/architecture/adr to infer conventions.
Boundary markers: No delimiters or ignore instructions are provided when reading existing files.
Capability inventory: The skill is permitted to use Read, Write, Edit, and Glob tools.
Sanitization: There is no automated sanitization of the content read from the repository.
Mitigation: The workflow includes a mandatory user review phase (Phase 4) where the final draft and destination path must be approved before execution.

adr-drafting