The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from technical change records and project metadata files, creating an attack surface for indirect prompt injection.\n
Ingestion points: The scripts tc_init.py, tc_status.py, and tc_update.py read content from files like CLAUDE.md, package.json, tc_registry.json, and individual tc_record.json files.\n
Boundary markers: The skill does not implement explicit delimiters or boundary markers when rendering record content for the agent's context.\n
Capability inventory: The agent has permissions to perform file system modifications, directory creation, and shell command execution for the provided Python scripts.\n
Sanitization: The scripts perform schema validation but do not sanitize or filter the natural language content within fields such as summary, notes, or handoff blocks before they are interpreted by the agent.

tc-tracker