The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a local shell script, lint-adr.sh, to perform static analysis on ADR files. The script is self-contained and utilizes standard Unix text-processing utilities (grep, bash, tr, cut) to verify section order, naming conventions, and status vocabulary without executing the content of the files or making network requests.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external inputs such as discussion transcripts and existing ADR documents to generate formatted output. This represents a functional surface for indirect prompt injection; however, the risk is mitigated by the skill's reliance on a strict markdown template and a validation script that enforces structure.
Ingestion points: Discussion transcripts, resolve notes, and existing ADR files provided in the user prompt.
Boundary markers: The instructions direct the agent to read source content fully before drafting, though explicit delimiter markers for untrusted input are not defined.
Capability inventory: File system read/write, local script execution (lint-adr.sh), and integration with the reversibility-rank skill.
Sanitization: The skill performs structural validation of the generated output via regex-based checks in the lint-adr.sh script.

adr-shape