Add on-demand GPT goal delegation to pstack - #356
Conversation
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes using high effort and found 3 potential issues.
Bugbot Autofix is ON, but it could not run because the branch was deleted or merged before autofix could start.
Reviewed by Cursor Bugbot for commit 9551634. Configure here.
|
|
||
| ## make it yours | ||
|
|
||
| Use [`/goal`](./skills/goal/SKILL.md) when setting a goal and delegating bounded tasks. Astra at medium effort owns integration and verification, Luna at max handles exploration and research, and Sol at high implements. An independent Astra review at xhigh runs only when needed. The skill checks model availability before dispatch and uses the current root session if it cannot change it. See [`/setup-pstack`](./skills/setup-pstack/SKILL.md) to override goal roles. |
There was a problem hiding this comment.
Inventory omits the goal skill
Medium Severity · Bugbot Rules
The new goal skill is missing from the README skills table. Goal with subagents also landed in the poteto-mode playbook list without a playbooks/ file or a matching README table row, so inventory surfaces no longer match the tree. Matching that entry then asks the agent to open a playbook file and copy its steps verbatim, which cannot succeed.
Additional Locations (2)
Triggered by learned rule: pstack inventory counts must match the skill/playbook tree
Reviewed by Cursor Bugbot for commit 9551634. Configure here.
| --- | ||
| name: goal | ||
| description: Run an explicit goal with on-demand subagents. Use when the user sets a goal and asks to delegate tasks, or requests the cost-efficient GPT agent tree. Own integration and verification at the root. | ||
| --- |
There was a problem hiding this comment.
Goal skill allows model auto-invocation
Medium Severity · Bugbot Rules
The new goal skill is a user-invoked coordinator and omits disable-model-invocation: true. Peers such as how, swarm, figure-it-out, and architect set that flag so the model does not auto-select them. Without it, ordinary goal or delegate phrasing can pull work onto the GPT role table and override other workflow defaults.
Triggered by learned rule: pstack disable-model-invocation: coordinators yes, setup-pstack no
Reviewed by Cursor Bugbot for commit 9551634. Configure here.
|
|
||
| # Setup pstack | ||
|
|
||
| For the cost-efficient GPT goal workflow, use the [goal skill](../goal/SKILL.md). Its table owns the default goal roles and reasoning efforts. An explicit request for that tree already selects those defaults; ask only about unavailable choices. Keep goal overrides separate from the other roles below. Store model and effort separately, for example `goal worker: gpt-5.6-sol` and `goal worker effort: high`. In Cursor, include those lines in the rule below. In other harnesses, use their existing always-loaded instruction file and name the target before editing. Preserve unrelated instructions. If no override is needed, the goal skill's defaults suffice. Root settings describe the desired session and do not change a running model. |
There was a problem hiding this comment.
Setup prose joins clauses with semicolon
Low Severity · Bugbot Rules
New setup-pstack prose joins two independent clauses with a semicolon in already selects those defaults; ask only about unavailable choices. Unslop punctuation in pstack/skills/ treats clause-joining semicolons as banned and splits them into a period plus a capital.
Triggered by learned rule: pstack skill prose follows unslop punctuation
Reviewed by Cursor Bugbot for commit 9551634. Configure here.


Adds an on-demand GPT delegation workflow for explicit goals. Astra at medium owns integration and verification, Luna at max handles bounded exploration and research, and Sol at high implements. Independent Astra review at xhigh runs when a concrete risk warrants it.
The goal skill checks available models and spawn arguments, keeps small work at the root, and requires verification of the combined result before completion. Poteto-mode routes goal delegation to it, and setup-pstack preserves separate model and effort overrides. Other workflows retain their defaults.
Validated the goal and setup skill files and checked four instruction scenarios with an independent Luna agent: small implementation, parallel investigation and research, unavailable models, and failed integration. This checks instruction behavior, not a full implementation run. Root model selection remains a session setting.
Note
Low Risk
Changes are agent instruction markdown only; they alter delegation behavior when users invoke
/goalor matching poteto-mode routes, with no application runtime code.Overview
Introduces
/goal, a new skill for explicit goals with on-demand subagent delegation (cost-efficient GPT role tree: Astra root at medium, Luna for exploration/research at max, Sol for implementation at high, optional Astra review at xhigh). The skill defines dispatch rules (model availability checks, harness-specific spawn args, root stays on the current session model), a seven-step runbook (scope, spawn only useful roles, root integration and verification, reviewer only for concrete risk), and overrides poteto-mode defaults like mandatory feature fan-out and orchestrate’s no-root-edits policy for that workflow.poteto-modenow points explicit goal delegation at the goal skill (subagents section + new Goal with subagents playbook entry).setup-pstackdocuments separategoalmodel/effort overrides inpstack-models.mdc, goal-only setup short-circuit, and preserving existing goal lines on rewrite. README documents when to use/goaland links to setup overrides.Reviewed by Cursor Bugbot for commit 9551634. Bugbot is set up for automated code reviews on this repo. Configure here.