Build project-ledger into a durable inventory system for projects, repositories, and idea vaults spread across multiple computers and storage roots.
The repo currently contains a working prototype scanner. Your job is to harden it into a maintainable toolchain that supports:
- repeated ingestion across multiple machines
- stable project identity across runs and devices
- manual curation where heuristics are insufficient
- mergeable machine-specific outputs
- human-readable and agent-readable artifacts
Read these files in this order before making non-trivial changes:
README.mdPRD.mdARCHITECTURE.mdSCHEMA.mdROADMAP.mdBACKLOG.mdRUNBOOK.md
As of March 8, 2026:
build_ledger.pyis the working prototype CLIledger_config.jsondefines the current scan rootstemplates/project-ledger.sidecar.example.jsondefines the sidecar patternprompts/session_end_prompt.mddefines the session-end ledger update prompt- the prototype emits CSV, JSON, and Markdown outputs
- unit tests exist but coverage is intentionally small
- Do not break the existing
python build_ledger.pyflow while refactoring. - Preserve deterministic outputs unless the schema/version explicitly changes.
- When changing schema, update:
SCHEMA.md- sidecar example template
- tests
- README and runbook if operator behavior changes
- When adding a new capability, add at least one automated test.
- Prefer standard library solutions unless a dependency is clearly justified.
- Generated artifacts in
output/are not canonical source files. - Keep sidecar files human-editable and conservative. Do not force complex manual workflows.
- Any work that changes session-end metadata expectations must also update
prompts/session_end_prompt.md.
python build_ledger.py
python -m unittest tests\test_build_ledger.pyIf you introduce new commands, document them in README.md and RUNBOOK.md.
A change is not done unless:
- the code path works locally
- tests pass
- docs reflect the new behavior
- schema changes are documented
- the handoff path for future agents remains clear
Parallel work is fine, but sequence dependencies matter:
- Schema hardening
- Package refactor and config validation
- Merge and multi-machine support
- Reporting and operator UX improvements
- Automation and optional UI layers
When you finish work in this repo, update relevant docs and, if appropriate, refresh any sample sidecar or output examples. Use prompts/session_end_prompt.md as the pattern for how downstream repos should be updated.