Skip to content

eval: guard shared workspace render ownership - #51

Merged
myobie merged 1 commit into
mainfrom
schickling-assistant/2026-07-30-shared-workspace-render-eval
Jul 31, 2026
Merged

eval: guard shared workspace render ownership#51
myobie merged 1 commit into
mainfrom
schickling-assistant/2026-07-30-shared-workspace-render-eval

Conversation

@schickling-assistant

@schickling-assistant schickling-assistant commented Jul 30, 2026

Copy link
Copy Markdown
Contributor

Paired with compoundingtech/st2#95 and merged runtime PR compoundingtech/st2#97.

Discriminator

Adds one model-free cell for two active declarations that share a workspace:

  • different effective bytes for .st2/PERSONA.md must fail validation
  • full materialization must fail before either owner writes
  • targeted materialization must not bypass the unselected sibling's claim
  • diagnostics must name the target and both owners
  • byte-identical shared claims remain valid and idempotent

The cell starts no PTY, Claude, or Codex session. It also updates the generated catalog, explicit model-free harness exclusion, and eval VRS mapping.

AGENT-SPEC.md remains pinned to its current accepted st2 commit in this PR; the new behavior should become normative there only when the repository-wide pin advances.

Current proof

Exact PR head: 1a73f98816cf2c48610bcef59f9aa0de80b8a35d.

Against current merged st2 c6846f6239329f0803142afc06c15a07b93937c1 (binary SHA256 2bba8d58be24250bc262f75f835ce2d780369add275774f3f2135c623d23d29c):

SCORE: 7 PASS / 0 FAIL / 7 gating judges
VERDICT: PASS

The exact head merges cleanly onto current evals main ffe2b14c785018cf365ea4ad87ef634029732b87. On that temporary merge, the complete repository-pinned model-free preflight passed with exact st2 9887b2842222def0838c2cd82e6c24c218f7efa6 / binary SHA256 d49d44fd4f3f6f655455c212353a469fefa956082bedf22163deb767d8a36a0d:

  • 73 KDL files parse
  • 46 maintained eval folders load; 9 nested catalogs validate strictly
  • model-free exclusion inventory matches exactly
  • fixture resets and harness contract pass
  • no-PII and generated CATALOG.md checks pass
  • git diff --check passes

No model/provider call was made.

Historical proof

Before runtime PR #97 merged, the original branch proof was also 7/7 against its then-exact runtime head d691d71. That result is historical; the current proof above supersedes it.

agent-session-id: a078daee-6f98-4916-91a8-d21291407789
agent-tool: Codex CLI
agent-tool-version: 0.145.0
agent-model: unknown
agent-runtime-profile: /nix/store/mnx8agbdq3wiyb6vz63lhgscgazkrn98-coding-agent-runtime-profile/share/coding-agents/profile.json
agent-skills-manifest: /nix/store/5r69m9k2llmri3na81518zx0a7y0d3cn-agent-skills-corpus/share/agent-skills/manifest.json
tooling-profile: dotfiles@0fb7e03

@myobie myobie left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Owner review GREEN: focused model-free cell 7/7 on current merged st2 c6846f6; clean merge onto evals main ffe2b14; complete repository-pinned 9887b28 model-free preflight, generated catalog, harness exclusions, fixture resets, no-PII, shell syntax, and diff check all pass. No provider call.

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.

2 participants