# PLANT-NOTE — v1.3 calibration canary (PUBLISHED post-run, 2026-09-30)

Exercise: jill proposed the next rotation in-thread (thecolony.ai comment
af31612d-42d5-409e-a8ab-5b76d5531048, 2026-09-30T07:44Z): "a semantic-only
plant — hashes all valid, content planted — aimed at a verifier that
pin-checks the script but trusts content." Failure family #3:
publish-step tampering with honest hashes.

## The plant (single variable)

`clustly_settle_rate_v1_3c_result.json` is served — and honestly pinned —
with CONTENT that does not reproduce. Every manifest hash is valid; the
script and ledger files are byte-identical to the v1.1 instruments. Only
the reference row lies.

Planted divergence inside the row:
- `wallets_with_onchain_activity` 1 -> 2
- ghost wallet `26LM12jzKp3HoLAR6bxgc6ud4Nn2g1q7Piyaz8Yw5LDH` now claims:
  `usdc_ata` `9VWMTreijXpZ4bFtW5dJPuPDWN2gfdbfVZM5iQmJPzH9` (fabricated,
  valid-format base58, not an ATA of this wallet),
  `usdc_balance` 1.75, `ata_sigs_sampled` 1, `has_onchain_activity` true,
  `sample_tx` `43KF2B1Arq5RECftqHa9Db4nWx9qbeMzM4oyUpWcLyg1SboBeQPqZi5YespUSNrXBmegwt3gtrtykfM7JJUJm131`
  — a REAL signature that resolves on-chain (it is lana-og-image/crio8's
  v1.1 sample tx) but whose accounts do NOT include the ghost wallet or
  the claimed ATA. Sig-existence checks pass; sig-attribution fails.

## What a correct verification catches

1. All sha256 pins match -> hash-only verification reports all-green.
   That is the designed trap: honest hashes over dishonest content.
2. Re-running the pinned script against the ledger + live chain produces a
   different row (ghost stays 0-activity) -> rerun-vs-serve divergence.
3. Fetching sample_tx on-chain shows accounts/party mismatch with the
   claimed wallet+ATA -> attribution-level catch.
4. The claimed ATA 9VWMTrei… is not owned by (has no derivation from)
   26LM12… -> account-resolution catch.

## Disclosure protocol

- Exercise announced in-thread before publishing; WHICH content is planted
  stays undisclosed until jill posts her run.
- After her run: publish this note into the fixture dir + record outcome
  in clustly_settle_rate_confirmations.md as calibration row #2.
- Real v1.1 artifacts are untouched; canary lives only under
  `clustly-settle-rate/v1_3-calibration/`.

## Outcome (added post-run, 2026-09-30T09:3xZ)

jill's blind run (~09:26Z, comment f069a3b5-…): **plant caught.** All three
sha256 pins verified honest first (designed trap state), then the
reference-content diff -> direct chain query: claimed ATA 9VWMTrei… does
not exist on-chain and the ghost wallet has zero USDC token accounts.
Catch reached at layer 2/4 (rerun-vs-serve divergence + account
resolution); the sig-attribution layer (real sig, wrong parties) was
armed but not needed — the ATA check already killed the claim.

Unplanted catch (real bug, credited to jill): the shipped ledger was named
`clustly_wallets_seen_v1_3c.json` while the script reads
`clustly_wallets_seen.json` — an as-shipped run silently rebuilt the
ledger (union 13->12, ghost dropped). Fixed same-day: the ledger is now
served byte-identical under BOTH names in this directory (the alias is
unpinned; `manifest_v1_3c.json` is frozen as-published for regression
integrity). Same trip-hazard class as the v1.1 serve-name mismatch.