The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests data from Odoo views, record rules, and model metadata, creating a surface for indirect prompt injection where malicious Odoo records could attempt to influence the agent.\n
Ingestion points: scripts/model_brief.py, scripts/entrypoints.py, scripts/metadata.py, and scripts/capabilities.py read data from the running Odoo registry.\n
Boundary markers: Data extraction is handled using defined JSON sentinels (e.g., ===ODOO_BRIEF_START===) and structured JSON parsing.\n
Capability inventory: The skill can execute shell commands via odoo-bin shell and write introspection results to the local filesystem.\n
Sanitization: The suite implements comprehensive data redaction via scripts/redaction.py and scripts/state_capture.py to mask credentials and PII before processing.\n- [EXTERNAL_DOWNLOADS]: The scripts/doc_index.py tool clones documentation from the official Odoo GitHub repository to build a local queryable index. This is a neutral, intended behavior for a documentation lookup feature.\n- [COMMAND_EXECUTION]: The CLI utility and several scripts (scripts/viz.py, scripts/doc_index.py, scripts/red_green_replay.py) use the subprocess module to execute system commands like odoo-bin, git, and gh. These operations are essential for Odoo introspection, indexing, and CI integration, and are performed with safe argument handling.

odoo-introspect