Skip to content

replication150.md

Latest commit

 

History

History
36 lines (26 loc) · 1.07 KB

replication150.md

File metadata and controls

36 lines (26 loc) · 1.07 KB
Raw

PsyClaw 150-paper replication pipeline

Canonical marker: <folderName>.psyclaw only (legacy Builder files are not the deliverable).

Commands

# unit/integration tests
python -m pytest webui/tests/replication150 -q

# one paper dry-run
python -m tools.replication150.cli one cat1_stroop --manifest E:/hermes_playground/psyclaw-vault/catalog/papers.json --dry-run

# batch dry-run
python -m tools.replication150.cli batch --manifest E:/hermes_playground/psyclaw-vault/catalog/papers.json --category 1 --dry-run

# cold backup (required before real vault writes)
python -m tools.replication150.cli backup --vault E:/hermes_playground/psyclaw-vault --out E:/backup/psyclaw_150_<timestamp>

Validation ladder

  1. Corpus integrity (50+50+50)
  2. PDF resolve + audit
  3. Method evidence
  4. Marker validate + compile
  5. WebUI open parity
  6. Protocol assertions
  7. Autopilot ×3
  8. four-file analysis pack per run
  9. Pilot/visual sample

Materials policy

  • Public/downloadable stimuli for replication: allowed.
  • Application/license gated corpora: list only, do not download.