Squad bootstrap fails on fresh repositories due to stale validator pin and genesis validation #2067

Summary

The Squad bootstrap workflow installed into a fresh consumer repository completed without creating its expected Cast pull request and research-proposals issue. Investigation found two issues in the shipped bootstrap/validator bundle:

1. workflows/squad-bootstrap.md pinned a SHA-256 that did not match the shipped workflows/shared/squad-cast-validator.mjs resource.
2. After correcting that pin, first-run validation treated the intentional absence of a committed .squad/casting/registry.json and history.json as an error. The agent spent its execution window investigating an unavoidable genesis-state diagnostic and eventually timed out.

Reproduction

Install and run the Squad bootstrap workflow in a fresh repository with no existing .squad/casting/registry.json or .squad/casting/history.json committed at HEAD.

The initial run may emit noop instead of the materialization safe output because the generated validator runner expects:

31e568ae4a0cc372f5b79d4b024ba8b7af1f38feac54034221fb203da9918ab4

while the installed shared/squad-cast-validator.mjs resource hashes to:

62fbf47b51639fd1878c143e5176ee3099e390065997411511e9d483d467bbce

After correcting the pin, a subsequent first-run bootstrap can still time out while repeatedly investigating this diagnostic:

registry base: committed registry/history pair is unavailable or malformed

No safe-output item is emitted, so the materialization job is skipped.

Verified fixes

A consumer-side patch successfully resolved the failure by:

• aligning the bootstrap validator integrity pin with the exact distributed validator bytes;
• changing committedRegistry() so:
  • neither committed file exists: valid genesis state, return null without an error;
  • only one committed file exists: reject as an incomplete committed pair;
  • both exist: parse and validate as before;
• raising the bootstrap timeout from 20 to 30 minutes as additional protection;
• regenerating integrity pins in both squad.md and squad-bootstrap.md;
• recompiling the affected workflow lock files with the matching gh-aw compiler version.

After these changes, bootstrap completed successfully and created both expected artifacts.

Proposed upstream resolution

1. Apply the genesis-aware committedRegistry() behavior to the canonical validator.
2. Regenerate every integrity pin that authenticates squad-cast-validator.mjs.
3. Recompile the affected workflow lock files with the repository-pinned gh-aw compiler.
4. Add regression coverage for:
   • a fresh repository with neither committed registry/history file;
   • a repository with only one of the two committed files;
   • an existing valid committed pair;
   • integrity pins matching the exact distributed validator bytes.
5. Consider retaining the 30-minute bootstrap timeout, although deterministic genesis handling should substantially reduce runtime.