Skip to content

SP-2a spec: the harness supervisor in fleetd, proved against harness-fake - #96

Merged
adbarc92 merged 3 commits into
mainfrom
docs/sp2a-harness-supervisor-spec
Oct 10, 2026
Merged

adbarc92 merged 3 commits into
mainfrom
docs/sp2a-harness-supervisor-spec

Conversation

@adbarc92

@adbarc92 adbarc92 commented Oct 1, 2026

Copy link
Copy Markdown
Owner

The SP-2a design, v4, as it stood on this branch on 2026-09-16 after three critique rounds (§10). No changes since. Docs only: one new file, docs/superpowers/specs/2026-09-16-sp2a-harness-supervisor-design.md.

What it settles

SP-2 is too large for one plan, so it is cut in three:

  • SP-2a (this spec): a shared protocol monitor; the harness registry and GET /harnesses; the harness supervisor (caps, gates, halt and resume, failures) behind an EvidenceVerifier seam; demo mode running through harness-fake.
  • SP-2b: the real evidence verifier and the fleet-runner crate.
  • SP-2c: harness-claude-docker extracted, real mode switched to it, the in-process driver deleted.

2a adds the harness path beside today's driver::run. Demo units move to it; real units stay where they are until 2c.

Admitted as written

Decided 2026-10-01, with two notes that do not change the text:

What follows

The implementation plan (#81), then the four build issues: #82, #83, #84, #85.

…s-fake

First of three SP-2 slices (parent: NEXUS swappable-harness design v0.2; ADR 0003 accepted).
Approved in brainstorm, then revised through three independent critique rounds.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
@adbarc92
adbarc92 merged commit b83e33e into main Oct 10, 2026
18 checks passed
@adbarc92
adbarc92 deleted the docs/sp2a-harness-supervisor-spec branch October 10, 2026 18:26
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant