Pending human review. Edit freely before implementing.
Phase 2 · Part of #2
User story
As the forge maintainer, I want a runnable Pi coding agent wired to GLM-5.2 for every role, with Playwright drivable from Bash, so that the Pi binding and the empirical validation have a real environment to run in instead of an assumed one.
Proposed behavior
Stand up the Pi coding agent with GLM-5.2 configured as the model for every role. Install and verify Playwright can be driven from Bash. Confirm git worktrees and the gh CLI both work inside that environment. This is pure environment provisioning, independent of the command prose, so it can proceed in parallel with #3. If GLM-5.2 or Playwright cannot be provisioned, that is recorded here and blocks the downstream issues, rather than being discovered mid-validation. The maintainer can reproduce the setup from written steps and get a green trivial run.
Acceptance criteria
- A Pi environment exists in which GLM-5.2 answers as the model on a trivial end-to-end invocation.
- Playwright can be driven from Bash in that environment (a trivial navigate/screenshot succeeds).
git worktree add and the gh CLI both succeed in that environment.- The setup is documented so it can be reproduced.
Out of scope
Grounding
- Pi coding agent: the target harness. EXISTS (external).
- GLM-5.2: the open-weight model for every role. EXISTS (external).
- Playwright: the browser automation driver for the
web evaluation surface. EXISTS (external).
- The
gh CLI usage forge depends on is harness-neutral: commands/planning.md lines 63-74 and 156-175; commands/building.md lines 58, 77, 118, 315. EXISTS.
- The provisioned, documented Pi environment. NEW.
Dependencies
Blocked by none. Blocks #5, #6.
Granular contract to be negotiated by generator/evaluator at /forge:building time and posted back to this issue as a comment.
Phase 2 · Part of #2
User story
As the forge maintainer, I want a runnable Pi coding agent wired to GLM-5.2 for every role, with Playwright drivable from Bash, so that the Pi binding and the empirical validation have a real environment to run in instead of an assumed one.
Proposed behavior
Stand up the Pi coding agent with GLM-5.2 configured as the model for every role. Install and verify Playwright can be driven from Bash. Confirm git worktrees and the gh CLI both work inside that environment. This is pure environment provisioning, independent of the command prose, so it can proceed in parallel with #3. If GLM-5.2 or Playwright cannot be provisioned, that is recorded here and blocks the downstream issues, rather than being discovered mid-validation. The maintainer can reproduce the setup from written steps and get a green trivial run.
Acceptance criteria
git worktree addand theghCLI both succeed in that environment.Out of scope
Grounding
webevaluation surface. EXISTS (external).ghCLI usage forge depends on is harness-neutral:commands/planning.mdlines 63-74 and 156-175;commands/building.mdlines 58, 77, 118, 315. EXISTS.Dependencies
Blocked by none. Blocks #5, #6.
Granular contract to be negotiated by generator/evaluator at /forge:building time and posted back to this issue as a comment.