Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
116 lines (77 loc) · 2.86 KB

CONTRIBUTING.md

File metadata and controls

116 lines (77 loc) · 2.86 KB
Raw

Contributing to LR-Agent

Thanks for considering a contribution.

LR-Agent is a research-engineering project. The most valuable contributions are not necessarily new features; counterexamples, reproducible failures, benchmark tasks and clearer evidence are especially useful.

Good contribution areas

ChronoForge

  • reproducible demo repositories;
  • retrospective time-split datasets;
  • future-scenario quality checks;
  • Chrono Tournament implementation;
  • maintenance-cost telemetry;
  • calibration metrics;
  • misleading Patch Life Report counterexamples.

Causal Genome

  • near-neighbor tasks where a learned strategy should not transfer;
  • Anti-Gene false-positive/false-negative cases;
  • contamination/genealogy experiments;
  • verifier coverage;
  • confidence calibration.

Runtime

  • Windows/Linux portability;
  • safer tool execution;
  • tests;
  • typed interfaces;
  • provider compatibility;
  • UI observability.

Before opening a PR

Run:

python -m compileall lr_agent
pytest

If you changed the Web UI, also make sure the JavaScript still parses and the relevant workflow passes.

Research claims

Please separate:

  • implemented behavior;
  • experimental result;
  • hypothesis;
  • novelty claim.

Do not use phrases like “first ever” without strong evidence.

A useful pattern is:

“In the public literature we searched, we did not find an equivalent end-to-end mechanism.”

If you find prior work that overlaps with a claim in docs/research/, opening an issue or PR to correct the record is welcome.

Pull requests

A good PR explains:

  1. what problem it addresses;
  2. what changed;
  3. what executable evidence supports the change;
  4. what could still be wrong;
  5. whether research docs need updating.

Keep unrelated refactors separate from behavior changes when practical.

Counterexamples are first-class contributions

If LR-Agent:

  • promotes a bad Forge winner;
  • learns a harmful Gene;
  • misses a contamination path;
  • produces a misleading ChronoForge survival report;
  • triggers a noisy Epistemic Tripwire;
  • fails to reproduce a claimed behavior;

please open an issue with the smallest reproducible example you can provide.

A failed research hypothesis is still useful.

Security

Do not open public issues containing:

  • API keys;
  • private repository code;
  • access tokens;
  • sensitive infrastructure details.

See SECURITY.md.

Style

  • Python 3.11+.
  • Prefer small, testable functions.
  • Keep safety boundaries explicit.
  • Avoid silent fallback when it would make a result look successful.
  • Preserve provenance for experimental evidence.

Authorship

Repository-facing project authorship is maintained by LLR6. Contributors are naturally credited through Git history and GitHub contribution records.

License

By contributing, you agree that your contribution may be distributed under the repository's MIT License.