What problem does it solve? Coding agents often invoke grep, find, or awk through bash instead of using the native tool equivalents, and it is hard to tell whether the cause is a missing capability or simple habit. This Skill audits all recorded agent sessions to classify every bash-based search call, so you can decide whether to fix tool code, fix the schema description, or leave behavior alone. ## Core Features & Use Cases - Session-wide tool usage audit: Parses every session JSONL log, extracting tool_calls and content-block tool_use shapes to count native tool calls versus bash invocations of grep, rg, find, awk, and sed. - Cause classification with tags: Labels each bash search call with tags like OUTSIDE_WS, PIPE, FLAG_V, FLAG_F, FIND, and PLAIN_INWS to separate real capability gaps from habit and legitimate workspace-confinement cases. - Gap remediation workflow: Guides implementing missing flags in both the rg-backed and pure-Rust code paths, adding unit tests, running cargo test and clippy, and enriching the schema description to surface existing capabilities. - Use Case: You notice your agent keeps running grep through bash. Run this audit across 661 sessions and discover that invert (-v) and word (-w) flags are real gaps to implement, while pipe-to-head usage is an awareness problem fixed by improving the tool's schema description. ## Quick Start Audit all my agent sessions and tell me why the agent uses grep via bash instead of the native grep tool.
