Skip to content

port pstack 0.15.42 - #69

Merged
caelaxie merged 1 commit into
mainfrom
cursor/port-pstack-1236-5196
Nov 16, 2026
Merged

caelaxie merged 1 commit into
mainfrom
cursor/port-pstack-1236-5196

Conversation

@caelaxie

Copy link
Copy Markdown
Contributor

Upstream

Ports cursor/plugins#1236 (bf0c7cd80abdbf7fe47b634978a7d1b4071f71a7, feat(pstack): new default models, one-file why and how, statechart fixes, sharpen and reflect text). That is the only pstack/ commit since the last ported upstream commit c742dd61706f. Version is 0.15.42 in plugin.json, .factory-plugin/plugin.json, and the README header.

docs/harness/claude-code.md is not in this repo, so that file was not added.

What landed

  • Role defaults move off Grok. Code work, prose, and judgment are Opus 5.5. The hardest changes are Fable 5.2. Panels are Fable and Opus. /setup-pstack writes a # defaults: 2 marker and replaces recognized old defaults on the next run.
  • how and why are one-file skills. Their references/ trees are deleted.
  • Statechart mermaid backend, run.ts, and the chart, cli, corpus, labels, and render tests, including drawn.ts and the Login chart. Those files match upstream byte for byte. The package name stays @droid-skill/statechart.
  • Sharpen, reflect, and the small model-default wording updates in the README, guide, playbooks, and the skills that name a default model.

Droid deviations

Task has no per-spawn model field, so the new slugs land on the two seat droids.

  • Code seat poteto-agent (and poteto-agent-readonly) is claude-opus-5-5 at xhigh. It was grok-4.7 at xhigh. This seat covers feature, refactoring, bug-fix, perf-issue, hillclimb, judgment and prose, both reflect lenses, and swarm workers.
  • Judgment seat poteto-judgment (and poteto-judgment-readonly) is claude-fable-5-2 at max. It covers hardest tasks and one panel runner. Prose no longer uses this seat, because judgment and prose is Opus.
  • Upstream panels are claude-fable-5-2-high, claude-opus-5-5-high-fast, and gemini-4-carbon-high. There is no Gemini seat. The default panel is the two seats, and a list entry that names any other model id is dropped. The README says the default panel is fable 5.2 / opus 5.5.
  • Upstream swarm workers, reflect tooling, verification, and the panels run at high, some with -fast. A seat has one reasoningEffort. The code seat stays xhigh and the judgment seat stays max. Setup never writes a -fast slug. Fable has no fast mode.
  • Arena's cross-judge treats Fable and the other Claude models as one vendor, so a Claude parent would get Gemini. With no Gemini seat, a Claude parent skips the judge and says so. A parent from another vendor gets the judgment seat, then the code seat.
  • how spawns subagent_type: "explorer". Upstream spawns generalPurpose with readonly: true and no model.
  • why does not spawn. It is the upstream one-file text, verbatim, and it is off the spawn list, so a subagent may run it inline. comment-sicko traces a claim or runs /why. recall spawns one writable worker per source, because a read-only subagent has no MCP access.
  • /setup-pstack writes the pstack-models section of ~/.factory/AGENTS.md, pins seats with personal droids, and uses AskUser. The section stores model ids without an effort suffix. It names grok-* once, to recognize an old default. In a panel list that is not the old Opus-plus-Grok pair, each old slug becomes claude-opus-5-5 (upstream writes claude-opus-5-5-high-fast).
  • Guide page 1 describes the two seat efforts (code xhigh, judgment max) rather than a per-role high and a Gemini cap.
  • Reflect's three lenses all default to the code seat. Upstream moved tooling from Grok onto Opus and left judgment on Opus, so both lenses are the same model. The effort split (xhigh vs high-fast) is the seat's single xhigh.

Harness

node scripts/validate-plugin.mjs: OK (61 skills, 5 droids).

bun run typecheck and bun run test in skills/poteto-mode/scripts: typecheck clean, 61 pass, 0 fail.

TLA2TOOLS=/tmp/tlc/tla2tools.jar npm test in skills/statechart/scripts (tla2tools.jar 1.7.4, checksum 936a262061c914694dfd669a543be24573c45d5aa0ff20a8b96b23d01e050e88): 373 pass, 0 fail, 0 skipped.

droid plugin marketplace add pstack-ports/droid-pstack#cursor/port-pstack-1236-5196
Successfully added marketplace: droid-pstack@cursor-port-pstack-1236-5196

droid plugin install pstack@droid-pstack@cursor-port-pstack-1236-5196 --scope user
Successfully installed pstack@droid-pstack@cursor-port-pstack-1236-5196 (38ce983)

droid plugin list
PLUGIN                                            STATUS   SOURCE  VERSION
pstack@droid-pstack@cursor-port-pstack-1236-5196  Enabled  user    38ce983d1f4d

Installed plugin.json version is 0.15.42. Cache root:

/home/ubuntu/.factory/plugins/cache/droid-pstack@cursor-port-pstack-1236-5196-bed404193dc9/pstack-7171b73fae68/38ce983d1f4d

FACTORY_API_KEY was not available. Each droid exec activated the skill, then exited 1 with this error:

Skill "<name>" activated
Error during droid execution: Authentication failed. Please log in using /login or set a valid FACTORY_API_KEY environment variable.

Ran that way for setup-pstack, how, why, statechart, sharpen, reflect, architect, arena, blast-radius, no-comments, poteto-help, poteto-mode, recall, swarm, and teach. Every installed SKILL.md under the cache root is byte-identical to the same path in 38ce983, as are the changed statechart scripts (backends/mermaid.ts, run.ts, test/drawn.ts, and the other updated tests) and the three Benny skill files.

Installed skills/how/SKILL.md:

---
name: how
description: "Use for \"how does X work\", code walkthroughs before changing something, and placement / ownership / layering questions (\"where should this live\", \"which package owns this\", \"is this the right layer\"). Explains subsystem architecture, runtime flow, onboarding mental models. Use why for motivation."
disable-model-invocation: true
---

# How

Explore the codebase to answer "how does X work?" questions.

If the scope is ambiguous, state your interpretation and explore. The user can redirect.

## 1. Explain

Spawn one Task subagent with `subagent_type: "explorer"` that explores and explains in one pass. Give it the user's question and this brief:

> You are writing an architectural explanation for a senior engineer. Explore the codebase and write an explanation that a senior engineer unfamiliar with this area could read and walk away with a solid mental model. They should understand the architecture well enough to start working in it confidently. You have read-only access to the codebase.

## 2. Present

Present the subagent's output. Light edits for clarity or context from the conversation are fine. Do not substantially rewrite it.

Installed skills/why/SKILL.md:

---
name: why
description: "Use for 'why does X work this way', 'why we picked Y', design rationale, regressions, postmortems, or data-backed thresholds. Returns a cited read on decisions and tradeoffs. Use how for runtime behavior."
disable-model-invocation: true
---

# Why

Answer from the record, not the code: read the commits and PR discussions that introduced the target and each later change, and other sources if the trail is thin. Cite what the authors wrote, mark your own inferences as inferences, and say what the record does not answer.

Installed skills/sharpen/SKILL.md:

---
name: sharpen
description: "Make a task deeper, not bigger, then rewrite the prompt. Use for /sharpen."
disable-model-invocation: true
---

I want a deeper version of this task, not a bigger one. Before rewriting it, do four things:

1. Say what a top professional in this field would do here that I'm not asking for.
2. Name the rawest material they would start from instead of what has already been digested for me.
3. Name the specific methods and tools they would use.
4. Say what, other than you or me, would show the result is right.

Then rewrite my prompt to ask for exactly that.

Have the rewritten prompt ask the agent to prototype any choice an experiment can settle and recommend an option based on what it observed. Leave product or preference choices that no experiment can settle for me to make.

Installed seat pins, from droids/poteto-agent.md and droids/poteto-judgment.md in that cache:

  • poteto-agent: model: claude-opus-5-5, reasoningEffort: xhigh
  • poteto-judgment: model: claude-fable-5-2, reasoningEffort: max
Open in Web Open in Cursor 

Bring in the new role defaults, one-file how and why, statechart fixes, and the sharpen and reflect text, translated onto the Droid seats.

Co-authored-by: Kaitian Xie <caelaxie@users.noreply.github.com>

Copy link
Copy Markdown
Contributor Author

Exact-head harness proof for 38ce983d1f4d267691a4cf575c63cf9aa3bd27eb, preserved from the cloud run's results in this PR body:

droid plugin marketplace add pstack-ports/droid-pstack#cursor/port-pstack-1236-5196
Successfully added marketplace: droid-pstack@cursor-port-pstack-1236-5196

droid plugin install pstack@droid-pstack@cursor-port-pstack-1236-5196 --scope user
Successfully installed pstack@droid-pstack@cursor-port-pstack-1236-5196 (38ce983)

droid plugin list
pstack@droid-pstack@cursor-port-pstack-1236-5196  Enabled  user  38ce983d1f4d

Installed plugin version: 0.15.42. The run reports every installed SKILL.md, the changed statechart scripts/tests, and deslop/control-cli/control-ui files byte-identical to that head. Installed seat pins: poteto-agent claude-opus-5-5 / xhigh; poteto-judgment claude-fable-5-2 / max. Installed excerpts and cache path are in the PR body.

Validation reported by the run:

  • node scripts/validate-plugin.mjs: OK, 61 skills / 5 droids.
  • poteto-mode scripts: typecheck clean; 61 tests passed, 0 failed.
  • statechart: 373 tests passed, 0 failed/skipped, using tla2tools 1.7.4.
  • CI on this head completed without failures.

Runtime limitation: droid exec activated setup-pstack, how, why, statechart, sharpen, reflect, architect, arena, blast-radius, no-comments, poteto-help, poteto-mode, recall, swarm and teach, but each exited 1 with authentication failure because FACTORY_API_KEY was unavailable. This proves installation and skill activation, NOT successful model execution or subagent behavior. Full runtime verification remains blocked on authentication. Nothing merged.

@caelaxie caelaxie left a comment

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

pass. ports cursor/plugins#1236 (bf0c7cd, pstack 0.15.42), the only pstack commit since c742dd. versions match in plugin.json, .factory-plugin/plugin.json, and the readme header.

minimum changes and parity hold: statechart mermaid/run/tests/drawn.ts/Login chart match upstream byte for byte; why is verbatim; how differs only by the sanctioned generalPurpose+readonly → explorer map. seat, panel, effort, and path translations follow COMPATIBILITY.md, including the seat default update (code opus 5.5 xhigh, judgment fable 5.2 max), dropping gemini, why off the spawn list, and the # defaults: 2 migration. no cursor-team-kit mention added. embedded deslop/control-cli/control-ui untouched.

layout is the droid skills/ + .factory-plugin layout. ci "Validate plugin" is green. harness proof is droid plugin marketplace add + droid plugin install + list on 38ce983, plus installed-skill inspect; droid exec died only on missing FACTORY_API_KEY, which is accepted here. no issue to link.

squash-merging.

@caelaxie
caelaxie merged commit 91d3cc8 into main Nov 16, 2026
1 check passed
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