# Verifier notes (in addition to the harness report)

Run: sj-harness 0.8.2, rootless podman 5.4.2, container network mode "none" (checked with podman inspect during the run), all 255 external files fetched outside the container and checked by size and SHA-256. Every declared result of C1-C8 landed within tolerance; verdict reproduced for all eight claims.

## Integrity flags the node raised
- Benford deviations in `gaia_rotmod.csv:source_id`, `neighbor_controls.csv:tic` and `prsa2022_tessebs.tsv:#`: these columns are catalog identifiers (Gaia DR3 source IDs, TIC IDs, row numbers), not measurements, so Benford's law does not apply to them.
- Benford deviation in `gaia_rotmod.csv:best_rotation_period`: the rows are rotation periods for a selected sample of Sun-like stars, a range spanning about one order of magnitude, where Benford conformance is not expected. Not a finding.
- 7 repeated rows in `prsa2022_tessebs.tsv` (a VizieR extract of the Prsa et al. 2022 TESS eclipsing-binary catalog): `code/select_sample.py` uses this file only to build a set of TIC IDs to exclude, so repeated rows cannot change the selection. The selection itself re-derives exactly (R11.selection_reproduced = true, 39,937 stars / 121,190 light curves).

## Hazard screen
Read the paper, claims, code and data tables. Public astronomical data (TESS/MAST, Gaia, TIC, VizieR) and synthetic injected flares; no weapons-relevant content, no private data or credentials, no copied third-party work beyond cited catalog extracts and an MIT-licensed response table listed in data/external.json. Verdict: none.

No instructions aimed at verifiers were found in the bundle (harness scan also found no hidden content). `code/run_search.py` contains download code (requests) for the original search, but `code/run` does not call it, and the container had no network.