john/claude-engineering-plugin
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: tune ce:ideate volume model and presentation format

Browse Source 
Reduce per-agent idea target from 10 to 7-8 based on real usage data
showing ideas 8-11 were speculative tail that rarely survived filtering.
This keeps the unique candidate pool manageable (~20-30 after dedup)
while preserving frame diversity across 4-6 agents. Also add scannable
overview line before detail blocks in Phase 4, and clarify foreground
dispatch and native tool usage in Phase 1.
This commit is contained in:
Trevin Chow
2026-03-16 21:07:21 -07:00
parent b762c7647c
commit 3023bfc8c1
1 changed files with 7 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
plugins/compound-engineering/skills/ce-ideate/SKILL.md
View File

@@ -76,7 +76,7 @@ Infer two things from the argument:
- **Volume override** - any hint that changes candidate or survivor counts - **Volume override** - any hint that changes candidate or survivor counts
Default volume: Default volume:
- each ideation sub-agent generates about 10 ideas (yielding 40-60 raw ideas across agents, ~30-50 after dedupe) - each ideation sub-agent generates about 7-8 ideas (yielding 30-40 raw ideas across agents, ~20-30 after dedupe)
- keep the top 5-7 survivors - keep the top 5-7 survivors
Honor clear overrides such as: Honor clear overrides such as:
@@ -89,13 +89,13 @@ Use reasonable interpretation rather than formal parsing.
### Phase 1: Codebase Scan ### Phase 1: Codebase Scan
Before generating ideas, gather codebase context. This phase should complete in under 2 minutes. Before generating ideas, gather codebase context.
Run two agents in parallel: Run two agents in parallel in the **foreground** (do not use background dispatch — the results are needed before proceeding):
1. **Quick context scan** — dispatch a general-purpose sub-agent with this prompt: 1. **Quick context scan** — dispatch a general-purpose sub-agent with this prompt:
> Read the project's CLAUDE.md (or AGENTS.md / README.md if CLAUDE.md is absent), then list the top-level directory structure using the native file-search tool. Return a concise summary (under 30 lines) covering: > Read the project's CLAUDE.md (or AGENTS.md / README.md if CLAUDE.md is absent), then discover the top-level directory layout using the native file-search/glob tool (e.g., `Glob` with pattern `*` or `*/*` in Claude Code). Return a concise summary (under 30 lines) covering:
> - project shape (language, framework, top-level directory layout) > - project shape (language, framework, top-level directory layout)
> - notable patterns or conventions > - notable patterns or conventions
> - obvious pain points or gaps > - obvious pain points or gaps
@@ -121,8 +121,8 @@ Do **not** do external research in v1.
Follow this mechanism exactly: Follow this mechanism exactly:
1. Generate the full candidate list before critiquing any idea. 1. Generate the full candidate list before critiquing any idea.
2. Each sub-agent targets about 10 ideas by default. With 4-6 agents this yields 40-60 raw ideas, which merge and dedupe to roughly 30-50 unique candidates. Adjust the per-agent target when volume overrides apply (e.g., "100 ideas" raises it, "top 3" may lower the survivor count instead). 2. Each sub-agent targets about 7-8 ideas by default. With 4-6 agents this yields 30-40 raw ideas, which merge and dedupe to roughly 20-30 unique candidates. Adjust the per-agent target when volume overrides apply (e.g., "100 ideas" raises it, "top 3" may lower the survivor count instead).
3. Push past the safe obvious layer. The first 10-15 ideas are often the least interesting. 3. Push past the safe obvious layer. Each agent's first few ideas tend to be obvious — push past them.
4. Ground every idea in the Phase 1 scan. 4. Ground every idea in the Phase 1 scan.
5. Use this prompting pattern as the backbone: 5. Use this prompting pattern as the backbone:
- first generate many ideas - first generate many ideas
@@ -132,7 +132,7 @@ Follow this mechanism exactly:
7. Give each ideation sub-agent the same: 7. Give each ideation sub-agent the same:
- grounding summary - grounding summary
- focus hint - focus hint
- per-agent volume target (~10 ideas by default) - per-agent volume target (~7-8 ideas by default)
- instruction to generate raw candidates only, not critique - instruction to generate raw candidates only, not critique
8. When using sub-agents, assign each one a different ideation frame as a **starting bias, not a constraint**. Prompt each agent to begin from its assigned perspective but follow any promising thread wherever it leads — cross-cutting ideas that span multiple frames are valuable, not out of scope. Good starting frames: 8. When using sub-agents, assign each one a different ideation frame as a **starting bias, not a constraint**. Prompt each agent to begin from its assigned perspective but follow any promising thread wherever it leads — cross-cutting ideas that span multiple frames are valuable, not out of scope. Good starting frames:
- user or operator pain and friction - user or operator pain and friction