It sits between Execute and eXamine in the APEX flow, scores itself against an exit threshold (>=90% proceed, 70-89% document gaps, <70% iterate), and persists its findings to .codex/apex/docs/elicit-{task-slug}.json so a later pass can diff against prior verdicts instead of restarting.
Elicitation Skill
Purpose
Enable expert agents to self-review and self-correct their code before external validation (sniper). Based on BMAD-METHOD's 75 elicitation techniques.
3 Execution Modes
Mode 1: MANUAL (default)
Expert presents 5 relevant techniques → User chooses → Expert applies
Mode 2: AUTO (--auto)
Expert auto-detects code type → Auto-selects techniques → Applies silently
Mode 3: SKIP (--skip)
Skip elicitation → Go directly to sniper validation
Quick Start
After Execute phase, expert runs:
# Manual mode (default)
> Apply elicitation skill
# Auto mode (no prompts)
> Apply elicitation skill --auto
# Skip self-review
> Apply elicitation skill --skip
Workflow Overview
┌─────────────────────────────────────────────────────────┐
│ ELICITATION WORKFLOW │
│ │
│ Step 0: Init → Load context │
│ Step 1: Analyze Code → Detect code type │
│ Step 2: Select → Choose techniques (or auto) │
│ Step 3: Apply Review → Execute techniques │
│ Step 4: Self-Correct → Fix own issues │
│ Step 5: Report → Summary before sniper │
└─────────────────────────────────────────────────────────┘
Auto-Detection Matrix
| Code Type Detected |
Auto-Selected Techniques |
| Authentication/Security |
Security Audit, OWASP Check, Input Validation |
| API Endpoints |
Error Handling, Type Coverage, API Contracts |
| Database/ORM |
N+1 Detection, Migration Safety, Data Integrity |
| UI Components |
Accessibility, Edge Cases, Loading States |
| Business Logic |
SOLID Compliance, Unit Test Coverage, Edge Cases |
| Refactoring |
Breaking Changes, Regression Analysis, Backward Compat |
| Performance Critical |
Profiling, Memory Analysis, Complexity Check |
Config/Docs/Plugin files (.md agents/skills, hooks.json, frontmatter YAML) |
CQ-01, DOC-01, INT-01 + validation by the strictest parser in the consumption chain (e.g. js-yaml strict for frontmatter, json.tool for JSON) -- never "looks well-formed" |
Technique Categories (12)
Full catalog: references/techniques-catalog.md
- Code Quality (7): Code review, Pattern detection, Complexity analysis...
- Security (7): OWASP audit, Input validation, Auth check...
- Performance (6): Profiling, N+1 detection, Memory analysis...
- Architecture (6): SOLID check, Dependency analysis, Coupling review...
- Testing (6): Edge cases, Boundary testing, Error paths...
- Documentation (6): API review, Comment check, Type coverage...
- UX (6): Accessibility, Error messages, Loading states...
- Data (6): Schema validation, Migration safety, Data integrity...
- Concurrency (6): Race conditions, Deadlock analysis, State sync...
- Integration (7): API contracts, Backward compat, Breaking changes...
- Observability (6): Logging, Metrics, Error tracking...
- Maintainability (6): Readability, Naming, File organization...
Total: 75 techniques
Integration with APEX
A-nalyze → P-lan → E-xecute → [ELICIT] → X-amine
│ │ │
▼ ▼ ▼
Expert Expert sniper
code self-review (final)
Benefits:
- Expert catches own mistakes before sniper
- Faster validation (less sniper corrections)
- Knowledge retention (expert learns from self-review)
Forbidden
- ❌ Skip init step (must load context)
- ❌ Apply techniques without understanding code type
- ❌ Self-correct without documenting changes
- ❌ Report without listing applied techniques
- ❌ Use techniques outside expertise domain
Exit Criteria (Step 5: Report)
Score = techniques applied / techniques selected, weighted by critical category (Security, Architecture).
| Score |
Status |
Action |
| ≥ 90% |
🟢 |
Proceed to sniper |
| 70-89% |
🟡 |
Document gaps, then proceed |
| < 70% |
🔴 |
Iterate before sniper |
Self-correction failure: If a self-correction breaks the code → revert that correction and keep the finding as a report item instead.
Artifact Contract
Step 5 persists .codex/apex/docs/elicit-{task-slug}.json so a later pass
diffs against prior verdicts instead of restarting from scratch. Full
contract, {task-slug} derivation, and JSON schema: references/artifact-contract.md.
Steps Reference
| Step |
File |
Purpose |
| 0 |
steps/step-00-init.md |
Load context, detect mode, load prior artifact if present |
| 1 |
steps/step-01-analyze-code.md |
Analyze written code |
| 2 |
steps/step-02-select-techniques.md |
Select techniques |
| 3 |
steps/step-03-apply-review.md |
Apply review |
| 4 |
steps/step-04-self-correct.md |
Self-correct |
| 5 |
steps/step-05-report.md |
Generate report, persist elicit-{task-slug}.json |
1---2name: elicitation-23description: Use when an expert agent self-reviews and self-corrects code after the Execute phase, before sniper validation (BMAD-METHOD elicitation techniques).4---56<objective>7Elicitation lets an expert agent self-review and self-correct its own code before external validation, drawing on 75 elicitation techniques across 12 categories (code quality, security, performance, architecture, testing, docs, UX, data, concurrency, integration, observability, maintainability) inspired by BMAD-METHOD. Three modes control how techniques are chosen: MANUAL (default, user picks from 5 presented options), AUTO (auto-detected and applied silently), and SKIP (bypass straight to sniper).89It sits between Execute and eXamine in the APEX flow, scores itself against an exit threshold (>=90% proceed, 70-89% document gaps, <70% iterate), and persists its findings to `.codex/apex/docs/elicit-{task-slug}.json` so a later pass can diff against prior verdicts instead of restarting.10</objective>1112# Elicitation Skill1314## Purpose1516Enable expert agents to **self-review and self-correct** their code before external validation (sniper). Based on BMAD-METHOD's 75 elicitation techniques.1718---1920## 3 Execution Modes2122### Mode 1: MANUAL (default)23```24Expert presents 5 relevant techniques → User chooses → Expert applies25```2627### Mode 2: AUTO (--auto)28```29Expert auto-detects code type → Auto-selects techniques → Applies silently30```3132### Mode 3: SKIP (--skip)33```34Skip elicitation → Go directly to sniper validation35```3637---3839## Quick Start4041**After Execute phase, expert runs:**42```bash43# Manual mode (default)44> Apply elicitation skill4546# Auto mode (no prompts)47> Apply elicitation skill --auto4849# Skip self-review50> Apply elicitation skill --skip51```5253---5455## Workflow Overview5657```58┌─────────────────────────────────────────────────────────┐59│ ELICITATION WORKFLOW │60│ │61│ Step 0: Init → Load context │62│ Step 1: Analyze Code → Detect code type │63│ Step 2: Select → Choose techniques (or auto) │64│ Step 3: Apply Review → Execute techniques │65│ Step 4: Self-Correct → Fix own issues │66│ Step 5: Report → Summary before sniper │67└─────────────────────────────────────────────────────────┘68```6970---7172## Auto-Detection Matrix7374| Code Type Detected | Auto-Selected Techniques |75|--------------------|--------------------------|76| Authentication/Security | Security Audit, OWASP Check, Input Validation |77| API Endpoints | Error Handling, Type Coverage, API Contracts |78| Database/ORM | N+1 Detection, Migration Safety, Data Integrity |79| UI Components | Accessibility, Edge Cases, Loading States |80| Business Logic | SOLID Compliance, Unit Test Coverage, Edge Cases |81| Refactoring | Breaking Changes, Regression Analysis, Backward Compat |82| Performance Critical | Profiling, Memory Analysis, Complexity Check |83| Config/Docs/Plugin files (`.md` agents/skills, `hooks.json`, frontmatter YAML) | CQ-01, DOC-01, INT-01 + validation by the **strictest parser in the consumption chain** (e.g. `js-yaml` strict for frontmatter, `json.tool` for JSON) -- never "looks well-formed" |8485---8687## Technique Categories (12)8889Full catalog: `references/techniques-catalog.md`90911. **Code Quality** (7): Code review, Pattern detection, Complexity analysis...922. **Security** (7): OWASP audit, Input validation, Auth check...933. **Performance** (6): Profiling, N+1 detection, Memory analysis...944. **Architecture** (6): SOLID check, Dependency analysis, Coupling review...955. **Testing** (6): Edge cases, Boundary testing, Error paths...966. **Documentation** (6): API review, Comment check, Type coverage...977. **UX** (6): Accessibility, Error messages, Loading states...988. **Data** (6): Schema validation, Migration safety, Data integrity...999. **Concurrency** (6): Race conditions, Deadlock analysis, State sync...10010. **Integration** (7): API contracts, Backward compat, Breaking changes...10111. **Observability** (6): Logging, Metrics, Error tracking...10212. **Maintainability** (6): Readability, Naming, File organization...103104**Total: 75 techniques**105106---107108## Integration with APEX109110```111A-nalyze → P-lan → E-xecute → [ELICIT] → X-amine112 │ │ │113 ▼ ▼ ▼114 Expert Expert sniper115 code self-review (final)116```117118**Benefits:**119- Expert catches own mistakes before sniper120- Faster validation (less sniper corrections)121- Knowledge retention (expert learns from self-review)122123---124125## Forbidden126127- ❌ Skip init step (must load context)128- ❌ Apply techniques without understanding code type129- ❌ Self-correct without documenting changes130- ❌ Report without listing applied techniques131- ❌ Use techniques outside expertise domain132133---134135## Exit Criteria (Step 5: Report)136137**Score** = techniques applied / techniques selected, weighted by critical category (Security, Architecture).138139| Score | Status | Action |140|-------|--------|--------|141| ≥ 90% | 🟢 | Proceed to sniper |142| 70-89% | 🟡 | Document gaps, then proceed |143| < 70% | 🔴 | Iterate before sniper |144145**Self-correction failure**: If a self-correction breaks the code → revert that correction and keep the finding as a report item instead.146147---148149## Artifact Contract150151Step 5 persists `.codex/apex/docs/elicit-{task-slug}.json` so a later pass152diffs against prior verdicts instead of restarting from scratch. Full153contract, `{task-slug}` derivation, and JSON schema: `references/artifact-contract.md`.154155---156157## Steps Reference158159| Step | File | Purpose |160|------|------|---------|161| 0 | `steps/step-00-init.md` | Load context, detect mode, load prior artifact if present |162| 1 | `steps/step-01-analyze-code.md` | Analyze written code |163| 2 | `steps/step-02-select-techniques.md` | Select techniques |164| 3 | `steps/step-03-apply-review.md` | Apply review |165| 4 | `steps/step-04-self-correct.md` | Self-correct |166| 5 | `steps/step-05-report.md` | Generate report, persist `elicit-{task-slug}.json` |
fusengine