History

pstack: default explorer roles to grok (#166)

lauren committed 3 months ago

Commit 18bbf7a1 parent 4a648804 files · +5 −5Browse files

Unified · Split

4 files
  1. .cursor-plugin/plugin.json +1 −1
  2. skills/how/SKILL.md +1 −1
  3. skills/setup-pstack/SKILL.md +2 −2
  4. skills/why/SKILL.md +1 −1
.cursor-plugin/plugin.json+1−1
@@ -1,7 +1,7 @@
1{1{
2 "name": "pstack",2 "name": "pstack",
3 "displayName": "pstack",3 "displayName": "pstack",
4 "version": "0.11.6",4 "version": "0.11.7",
5 "description": "if you want to go fast, go deep first. pstack helps you write less, but higher quality code. rigorous agent workflows you can parallelize with confidence.",5 "description": "if you want to go fast, go deep first. pstack helps you write less, but higher quality code. rigorous agent workflows you can parallelize with confidence.",
6 "author": {6 "author": {
7 "name": "Lauren Tan"7 "name": "Lauren Tan"
skills/how/SKILL.md+1−1
@@ -45,7 +45,7 @@ The right decomposition depends on the question. Use your judgment. Narrow quest · 44 unmodified lines
45Spawn all explorers in a single message:45Spawn all explorers in a single message:
46 46
47- `subagent_type`: `generalPurpose`47- `subagent_type`: `generalPurpose`
48- `model`: resolve the configured `how explorer` role per `~/.cursor/rules/pstack-models.mdc`. Pass a real slug as `model`. Omit `model` for `inherit-parent`/`auto`. If the role line is absent, default to `gpt-5.6-sol-max`.48- `model`: resolve the configured `how explorer` role per `~/.cursor/rules/pstack-models.mdc`. Pass a real slug as `model`. Omit `model` for `inherit-parent`/`auto`. If the role line is absent, default to `grok-4.5-fast-xhigh`.
49- `readonly`: `true`49- `readonly`: `true`
50 50
51Each explorer gets the same base prompt from `references/explorer-prompt.md` plus a specific exploration angle naming its slice. Each explorer should:51Each explorer gets the same base prompt from `references/explorer-prompt.md` plus a specific exploration angle naming its slice. Each explorer should:
skills/setup-pstack/SKILL.md+2−2
@@ -48,10 +48,10 @@ perf-issue: gpt-5.6-sol-max · 47 unmodified lines
48hillclimb: gpt-5.6-sol-max48hillclimb: gpt-5.6-sol-max
49judgment and prose: claude-fable-5-thinking-max49judgment and prose: claude-fable-5-thinking-max
50hardest tasks: claude-fable-5-thinking-max50hardest tasks: claude-fable-5-thinking-max
51how explorer: gpt-5.6-sol-max51how explorer: grok-4.5-fast-xhigh
52how explainer: claude-fable-5-thinking-max52how explainer: claude-fable-5-thinking-max
53how critics: claude-fable-5-thinking-max, gpt-5.6-sol-max, grok-4.5-fast-xhigh53how critics: claude-fable-5-thinking-max, gpt-5.6-sol-max, grok-4.5-fast-xhigh
54why investigators: gpt-5.6-sol-max54why investigators: grok-4.5-fast-xhigh
55why synthesizer: claude-fable-5-thinking-max55why synthesizer: claude-fable-5-thinking-max
56reflect tooling: gpt-5.6-sol-max56reflect tooling: gpt-5.6-sol-max
57reflect judgment, divergent, synthesizer: claude-fable-5-thinking-max57reflect judgment, divergent, synthesizer: claude-fable-5-thinking-max
skills/why/SKILL.md+1−1
@@ -117,7 +117,7 @@ Launch all matching investigators in a single message so they run concurrently. · 116 unmodified lines
117 117
118Subagent config (each):118Subagent config (each):
119- `subagent_type`: `generalPurpose`119- `subagent_type`: `generalPurpose`
120- `model`: resolve the configured `why investigators` role per `~/.cursor/rules/pstack-models.mdc`. Pass a real slug as `model`. Omit `model` for `inherit-parent`/`auto`. If the role line is absent, default to `gpt-5.6-sol-max`.120- `model`: resolve the configured `why investigators` role per `~/.cursor/rules/pstack-models.mdc`. Pass a real slug as `model`. Omit `model` for `inherit-parent`/`auto`. If the role line is absent, default to `grok-4.5-fast-xhigh`.
121- `readonly`: `false` (agent mode). **Do not use readonly/Ask mode.** It strips MCP access, which disables MCP-backed investigators entirely. The source control investigator would be safe in readonly, but keep modes uniform. Investigators still shouldn't write anything. That's a posture, not a sandbox.121- `readonly`: `false` (agent mode). **Do not use readonly/Ask mode.** It strips MCP access, which disables MCP-backed investigators entirely. The source control investigator would be safe in readonly, but keep modes uniform. Investigators still shouldn't write anything. That's a posture, not a sandbox.
122 122
123Each investigator gets:123Each investigator gets: