Work / Project 02
AI Agent Data Mapping & Secret Detection
Find agent logs, memory, configuration, and saved credentials through a read-only workflow with masked model inputs.
← WorkLocal Python workflow
Use case
Agent histories, configuration, and databases can hold useful project context alongside credentials.
How it works
A two-pass inventory and indexing pipeline. Deterministic scripts scan and mask sensitive text before model-assisted labeling. Findings retain file and line references.
Inspect the implementation
The source repository includes the inventory, masking, sampling, batching, merge, and leak-checking workflow, with documented fixture-based verification.
Scope
The mapping workflow reads a local data-store copy. Unreadable files are recorded with a reason, and the label set is reviewed before the full indexing pass.
Source & documentation
Three pipeline steps
Pass 1 inventories, scans, masks, samples, and batches. Pass 2 labels masked text and stops for human label review. Pass 3 validates locators, merges results, and checks outputs for leaks.
Review procedure
python -m pytest -q
Content reviewed against repository documentation on November 5, 2026.