release: sync Cursor pstack 0.14.7

[ericlitman]

Closes #43

What changed

• Port Cursor pstack 0.14.6's forge-neutral pull request playbooks while preserving Open Pstack's Claude Code and Codex mappings.
• Add schema-first TypeScript boundary guidance and .ts and .tsx automatic-loading scope for Claude Code, with explicit by-name use on Codex.
• Sync the release record to Cursor pstack 0.14.7, publish the exact upstream logo through the Codex manifest, and record the deliberate upstream exclusions.
• Keep bug-fix, perf-issue, and hillclimb on codex:gpt-5.6-sol@max.

Verification

• Bun tests, strict typecheck, static invariants, and plugin validation pass.
• The exact candidate's affected plugin package is installed in every affected harness.
• The changed behavior passes from each real user surface.
• The installed version, action, and observed result appear below.

Exact candidate: 99ace9ddfad6cf0bf4dec426de520f1295c7d605 on base f466256cc684f7cd44ee7e8eaeea2e3904921d84.

The only change from the fully installed and live-tested d140cda4fbdce0c093b34473e587c89e90608b94 candidate is the factual release header in root docs/reference.md. The installable plugin tree and all runtime behavior are byte-identical, so the live and performance evidence below applies unchanged to the exact head.

Local gates:

• All 158 Bun tests pass with 657 expectations.
• Strict typecheck, static invariants, four manifest parses, Claude plugin validation, marketplace validation, and git diff --check pass.
• Each of the three ordered commit boundaries passes its required gates.
• README-UPSTREAM.md matches efa2a53:pstack/README.md byte for byte.
• plugins/pstack/assets/logo.png matches efa2a53:pstack/assets/logo.png byte for byte at SHA-256 f18b7225edb7c5d93aa533856440007a08da4d7035693c13aef906b28bf57acf.

Claude Code 2.1.258 with Open Pstack 1.3.0 installed from the candidate's byte-identical plugin tree:

• Lane 1. The trunk response declares check mode. The head response declares check mode, selects GitHub through gh, and records that Origin is unavailable.
• Lane 2. /pstack:poteto-mode land the stack merged a real two-PR disposable stack bottom first, retargeted the child to main, reran CI, then merged the child. No gt command ran. The final tree matched the verified stack tip.
• Lane 3. Amending the child changed its patch ID and triggered fresh codex:gpt-5.6-sol@max verification. The later rebase kept the patch ID, kept the fresh verdict, reran CI, and landed bottom first. The final tree matched the verified tip.
• Lane 6. The exact-head review proposed a Zod schema, derived the type with z.infer, and cited "schemas before guards." The trunk model also knew the Zod pattern, so the negative baseline did not distinguish the versions. The requested head behavior passed on an identical repeat, and no source change was made for this model-output limitation.
• Lane 7. pstack:typescript-best-practices loaded through both the Skill tool and the slash command in a directory with no TypeScript file.
• Lane 9. pstack:how, pstack:why, and pstack:unslop loaded. The deliberately excluded pstack:make-bot-ui returned "Unknown skill."
• Lane 10. A fresh marketplace install listed Open Pstack 1.3.0. /pstack:poteto-mode fix the failing test delegated exactly once to codex:gpt-5.6-sol@max, with no timeout or fallback, and changed the fixture from one failure to one pass.

Codex with Open Pstack 1.3.0 installed from the candidate's byte-identical plugin tree:

• Lane 4. The marketplace checkout and installed cache match the candidate plugin tree. codex plugin list resolves the 1254 by 1254 logo. The session first declared check mode and GitHub through gh, with Origin unavailable, then read PR release: sync Cursor pstack 0.14.7 #44 without changing it.
• Lane 5. The shipping statement selected gh, described bottom-up one-at-a-time landing, and contained no gt command.
• Lane 8. The TypeScript review proposed a Zod object schema, derived the type with z.infer, and replaced the duplicate hand-written guard with schema-owned validation.

Performance:

• Claude Code ran five alternating trunk and head pairs. Trunk median was 2.94 seconds. Head median was 2.86 seconds, or 97.28 percent of trunk.
• Codex ran five alternating trunk and head pairs. Trunk median was 4.359 seconds. Head median was 4.084 seconds, or 93.69 percent of trunk.
• Every measured turn returned ok. Neither runner used a timeout or fallback. Both head medians stayed below the 120 percent limit.

Review status:

• Factory approved the full candidate at d140cda in run a637d820-8698-4922-9014-3ddade3bd8f1 with quorum 3 and zero findings.
• Factory approved the one-line d140cda..99ace9d documentation delta in run fa61bd85-4e19-4fdf-9491-2f6b753ecdb1 with quorum 3 and zero findings.
• Open SWE's first review found three potential issues. The stale reference header was fixed. The requested legacy Graphite migration and merge-group history tracker were dismissed as broader queue features outside this upstream sync.
• Open SWE re-reviewed exact head 99ace9d in run 01a0697e-b3a1-7fb1-96ab-af40fc1882c4. The exact-head check passed, no new finding was added, and all review threads are resolved.

Candidate identity:

• Diff SHA-256: ed7889a02cca4f58857248eb3d2343680c540639dbaf2223c0ba88a23a12efd8.
• Stable patch ID: 339b46f46a7c8d6fdb52d1c2024f45840fdafba2.
• Commit 1: e64b6f16f7f6f95927efee8dd7bf139ca0cee35c.
• Commit 2: 4bc46f4649c87de5490045592b10d2c225d39312.
• Commit 3: 99ace9ddfad6cf0bf4dec426de520f1295c7d605.

[greptile-apps]

Greptile Summary

This release synchronizes Open Pstack with Cursor pstack 0.14.7 while retaining its Claude Code and Codex adaptations.

• Reworks pull-request playbooks around forge-neutral, bottom-up stack delivery and guarded rebases.
• Adds schema-first TypeScript guidance and host-specific skill-loading metadata.
• Updates plugin manifests, release documentation, provenance records, tests, and the Codex logo asset.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains.

Important Files Changed

Filename	Overview
plugins/pstack/skills/poteto-mode/playbooks/shipping.md	Replaces the prior Graphite-oriented delivery flow with guarded GitHub/Origin stack landing, current-head checks, and stable patch-ID verdict handling.
plugins/pstack/skills/poteto-mode/playbooks/multi-phase-plan.md	Expands generated-plan guidance with trunk-versus-head regression evidence and dual-sided performance requirements.
plugins/pstack/skills/typescript-best-practices/SKILL.md	Adds TypeScript and TSX loading scope plus schema-first runtime-boundary guidance shared across supported hosts.
plugins/pstack/.codex-plugin/plugin.json	Updates the Codex plugin release metadata and exposes the synchronized upstream logo.
tests/skill-collision-repro.sh	Updates static plugin invariants for the new release metadata, TypeScript scope, model mappings, and asset registration.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  U[Cursor pstack 0.14.7] --> P[Open Pstack port]
  P --> W[Forge-neutral PR playbooks]
  P --> T[Schema-first TypeScript guidance]
  P --> M[Claude and Codex manifests]
  W --> C[Claude Code]
  W --> X[Codex]
  T --> C
  T --> X
  M --> C
  M --> X

Loading

_{Reviews (2): Last reviewed commit: "release: Open Pstack 1.3.0 at Cursor pst..." | Re-trigger Greptile}

[ericlitman]

/panel rerun

[ericlitman]

@openswe Please run Open SWE Review on this pull request at exact head e8315b1. Review only. Do not change the branch or merge. Limit findings to merge-blocking defects in the upstream integration. Ignore style nits and unrelated cleanup.

[openswebot]

Open SWE Review found 3 potential issues.

Open in Web

[mergify]

Tick the box to add this pull request to the merge queue (same as @mergifyio queue).

• Queue this pull request