The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection because it reads and processes content from the codebase, existing documentation, and configuration files to infer and generate architectural decisions without utilizing boundary markers or sanitization techniques.
Ingestion points: Codebase scanning for framework and infrastructure choices (Step 3.1), reading configuration files and code comments (Step 2.1), and scanning existing ADR markdown files (Step 1.3).
Boundary markers: Absent; the skill does not use delimiters to isolate ingested data from agent instructions.
Capability inventory: The skill is capable of writing markdown files to the local file system (docs and index) and logging telemetry to the user home directory.
Sanitization: No explicit escaping or validation is performed on the data read from the codebase before it is incorporated into the ADR documentation.
[NO_CODE]: The skill is composed exclusively of natural language instructions in markdown format and does not include any executable Python, Node.js, or shell scripts.

adr