The Agent Skills Directory

Standard Command Execution: The skill uses shell commands to inspect target code and apply updates to findings using tools like jq. These operations are limited to the user-specified target_root or the skill's own workspace/ directory, following standard practices for local security analysis agents.
File System Interaction: It reads codebase files and writes structured data to workspace/findings/ and workspace/learnings.jsonl. These actions are consistent with its documented purpose of tracking and categorizing security findings.
Snapshot Integrity (Safety Feature): The skill includes a 'sentinel check' to verify that it is analyzing the correct version of the source code (via .mantis_snapshot_id), which prevents incorrect assessments due to code changes (drift).

mantis-critic