The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local setup scripts and git commands to facilitate environment discovery and document management.
Evidence: Executes scripts/bash/setup-team-constitution.sh (or the PowerShell equivalent) to resolve repository paths and detect if a constitution already exists.
Evidence: Utilizes git add and git commit to version-control the changes made to the team constitution file.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface where it reads external documentation to inform its principle proposals.
Ingestion points: Reads content from CDR.md, AGENTS.md, and various rule files within the team directives directory.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are used when processing the ingested files.
Capability inventory: The skill possesses file-writing capabilities, shell script execution, and git command execution.
Sanitization: There is no explicit sanitization of the content retrieved from the team directives files before it is processed by the model.
Mitigation: The design incorporates a mandatory interactive review process, ensuring a human-in-the-loop approves, edits, or rejects every suggested principle before any file writes or commits occur.

team-constitution