The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user-editable project files to generate status reports, which is a potential surface for indirect prompt injection.\n
Ingestion points: Reads from PROJECT.md, STATUS.md, CLAIMS.md, DATA_DICTIONARY.md, ANALYSIS_PLAN.md, and REVIEW_LOG.md in scripts/init_project.py.\n
Boundary markers: Absent; data from these files is interpolated into status outputs without specific delimiters or instructions to ignore embedded content.\n
Capability inventory: The skill has file system write capabilities (Path.write_text in init_project.py) and performs network API calls to Zotero.\n
Sanitization: Absent; content from markdown memory files is processed as plain text and included in agent outputs.\n- [COMMAND_EXECUTION]: The skill executes an internal validation script using safe parameters.\n
Evidence: subprocess.run([sys.executable, str(validator), ...]) in scripts/migrate_project_to_ssot.py.\n
Note: This implementation avoids shell injection by passing arguments as a list and using the current system executable.\n- [EXTERNAL_DOWNLOADS]: The skill utilizes the pyzotero library for research integration.\n
Evidence: Import of pyzotero in scripts/init_project.py.\n
Note: Zotero is a well-known service for researchers, and this library is a standard tool for its API.

manage-project