etanhey/research-prompt-quality
skills/golem-powers/_archive/research-prompt-quality/SKILL.md
Mandatory pre-flight gate before any deep-research prompt ships. Three gates: CHECK-FIRST (non-redundancy), GROUND (Drive refs + current-usage examples + prior-research stance), emit-only-if-pass. Use when writing deep research prompts, Claude Desktop research prompts, deciding should we research, or proposing research. Triggers: 'deep research', 'research prompt', 'should we research', 'propose research'. NOT for executing research — use /research, /claude-desktop-research, or /gemini-research.
$ install --global
skillsauthnpx skillsauth add etanhey/golems research-prompt-qualityInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: May 30, 2026, 2:58 AM37.1s10 files scanned
SKILL.md
- name:
- research-prompt-quality
- description:
- Mandatory pre-flight gate before any deep-research prompt ships. Three gates: CHECK-FIRST (non-redundancy), GROUND (Drive refs + current-usage examples + prior-research stance), emit-only-if-pass. Use when writing deep research prompts, Claude Desktop research prompts, deciding should we research, or proposing research. Triggers: 'deep research', 'research prompt', 'should we research', 'propose research'. NOT for executing research — use /research, /claude-desktop-research, or /gemini-research.
Research Prompt Quality — Pre-Flight Gate
Stop flat, redundant deep-research prompts before they ship. Research execution skills run after this gate passes.
When to run
Run before writing or pasting any deep-research prompt (Claude Desktop, Gemini Deep Research, or /research --deep). If the gate fails, do not emit a research prompt — output the stop reason and route to engineering or plan work instead.
The three gates (all must pass)
Gate 1 — CHECK-FIRST (non-redundancy)
Search existing work before proposing new research.
skills/golem-powers/research-prompt-quality/scripts/check-first.sh "<topic keywords>"
Sources scanned:
Brain Drive/03_RESEARCH/(when mounted)- Every
~/Gits/*/docs.local/research/ - Every
~/Gits/*/docs.local/plans/ - Every
~/Gits/*/docs.local/decisions/
On exit 1: print ALREADY RESEARCHED → <paths> and STOP. This is engineering / plan execution, not research.
Canonical failure: proposing "RRF ranking" deep-research when ≥6 prior artifacts already exist (see evals/fixtures/neg-2-redundant-rrf.md).
Gate 2 — GROUND
Every research prompt MUST include:
- Drive folder refs — relevant
Brain Drive/03_RESEARCH/...paths (or documented folder IDs from boot grounding docs). - ≥1 concrete current-usage example — real code, config, or file path from the repo (not generic "we use agents").
- Prior-research stance — explicit BUILD-ON / VALIDATE / REFUTE for each prior artifact; never restart from zero.
Use references/ground-template.md as the required structure. Fold grounding bundles (e.g. MCL-RESEARCH-GROUNDING.md) into the prompt; do not treat the bundle as the prompt itself.
Gate 3 — Emit only if 1+2 pass
If CHECK-FIRST passes and GROUND is satisfied, emit the self-contained research prompt (reuse /claude-desktop-research format). Otherwise output only the stop message or grounding gap list.
Workflow
1. CHECK-FIRST → scripts/check-first.sh "<keywords>"
2. If hits → STOP ("ALREADY DONE → paths → engineering")
3. Gather ground → read repo paths, Drive folders, prior research files
4. Draft → references/ground-template.md
5. Score → scripts/score-research-prompt.py <draft.md> (target ≥8/10)
6. Ship prompt → hand to /claude-desktop-research or /research
Static quality bar
Run scripts/score-research-prompt.py on the draft. RED gate for eval fixtures: negative prompts ≤4/10, grounded prompts ≥8/10. See EVAL.md for rubric and fixture scores.
Integration
/research— run this gate before--deepor external research dispatch./claude-desktop-research— run this gate before writing the paste-ready prompt file.
Anti-patterns (from gen-10, 2026-05-29)
| Bad | Good |
|-----|------|
| Flat MCL prompt with no Drive refs or repo examples | Grounded prompt + MCL-RESEARCH-GROUNDING.md folded in |
| "Deep-research RRF fusion" when 6+ prior docs exist | ALREADY RESEARCHED → <paths> → engineering |
| Generic "multi-agent ecosystem" without send_input / cmux paths | Cite cmux/SKILL.md, orc/workflows/fact-propagation.md, etc. |
| Restart prior art from scratch | BUILD-ON / VALIDATE / REFUTE per prior file |
Eval
See EVAL.md — four real fixtures, literal scorer stdout, with-skill vs without-skill delta target ≥+30%.
Related Skills
etanhey/phoenix-human-view
tools
VerifiedTrustedCommunity
The human-eval UX contract for Phoenix views: turn-by-turn scrollable replay (not a scorecard), hide-but-copyable IDs, collapsed thinking, identity chips, tool filters, tiny frozen starter datasets, mark-wrong-in-thread, mobile-first. Use when: building or reviewing ANY Phoenix/eval view, annotation UI, session replay, or human-grading surface. Triggers: phoenix view, eval UI, annotation view, session replay, human eval UX, grading interface. NOT for: Phoenix data pipelines/ingest (capture scripts have their own specs).
3SKILL.mdUpdated Jun 7, 2026
etanhey/mac-systems
tools
VerifiedTrustedCommunity
macOS systems specialist — AppKit NSPanel architecture, launchd services, socket activation, MCP bridge resilience, syspolicyd, and high-frequency SwiftUI dashboards. Use when building menu-bar apps, LaunchAgents, debugging syspolicyd/Gatekeeper/TCC, resilient UDS/MCP bridges, or SwiftUI dashboards at 10Hz+.
3SKILL.mdUpdated Jun 7, 2026
etanhey/judge-fleet
development
VerifiedTrustedCommunity
Bulk LLM-judging protocol for fleet-dispatched verdict runs (KG cluster, eval harness). Use when: dispatching or running judge workers (J1/J2/RT), planning bulk-apply from verdict JSONL, or triaging evidence_degraded outputs. Triggers: judge fleet, bulk judge, R3 verdicts, kg-judge, RT gate, evidence_degraded. NOT for: single-item code review, Phoenix view UX (use phoenix-human-view), or non-judge eval pipelines.
3SKILL.mdUpdated Jun 7, 2026
etanhey/fleet-wrap
development
VerifiedTrustedCommunity
Quiet-down protocol for sprint close: when the fleet wraps, delete ALL polling crons and monitors, send ONE final dashboard + ONE message, then go SILENT. Use when: fleet wraps, all workers done, overnight queue exhausted, sprint close, Etan asleep/away with nothing approved left. Triggers: fleet wrap, wrap the fleet, stand down, going quiet, sprint close. NOT for: mid-sprint monitoring (keep your loops), spawning a successor (use /session-handoff first).
3SKILL.mdUpdated Jun 7, 2026