Agent-Skills.md

Agent Skills: Discovery — Phase 0 Validation Gate

Phase 0 validation gate — problem framing, competitive analysis, feasibility, and a go/no-go checkpoint before committing to a feature. Use when evaluating whether an idea is worth building, before /story or /design. Trigger on "is this worth building", "validate this idea", "discovery", "should we build".

UncategorizedID: matteocervelli/llms/discovery

Repository

matteocervelli/llms
matteocervelliLicense: MIT
252

Install this agent skill to your local

pnpm dlx add-skill https://github.com/matteocervelli/llms/tree/HEAD/.codex/skills/discovery

Skill Files

Browse the full folder contents for discovery.

Download Skill

Loading file tree…

.codex/skills/discovery/SKILL.md

Skill Metadata

Name
discovery
Description
Phase 0 validation gate — problem framing, competitive analysis, feasibility, and a go/no-go checkpoint before committing to a feature. Use when evaluating whether an idea is worth building, before /story or /design. Trigger on "is this worth building", "validate this idea", "discovery", "should we build".

Discovery — Phase 0 Validation Gate

Structured Discovery & Validation before committing to a feature. Prevents building unvalidated ideas by gating entry into the story and design pipeline with four sequential modes.

Usage

/discovery                   # Show overview of all 4 modes
/discovery problem           # Problem framing: who has it, how urgent, what evidence
/discovery competitive       # Competitive landscape: what exists, differentiators, moat
/discovery feasibility       # Feasibility: technical, economic, time constraints
/discovery checkpoint        # Go/no-go gate: 5-criteria verdict before /story create

Workflow

Step 1: Determine Mode

Parse $ARGUMENTS for explicit mode. If no arguments provided, show overview:

no args         → read templates/overview.md
problem         → read templates/problem/summary.md
competitive     → read templates/competitive/summary.md
feasibility     → read templates/feasibility/summary.md
checkpoint      → read templates/checkpoint/summary.md
<anything else> → show usage error + list valid modes + read templates/overview.md

Step 2: Present Guidance (Progressive Disclosure)

Level 1 — Summary (default, ~20 lines): Read and present the summary template. This is the entry point — frames the questions to answer.

Level 2 — Patterns (on request, ~60 lines): When the user asks for more depth, "show me patterns", "how do I score this", or "give me examples" — read the patterns template.

Level 3 — Full Analysis (on request): No template — dive into specific questions with the user based on context provided.

Step 3: Run Modes Sequentially

Recommended sequence: problem → competitive → feasibility → checkpoint

Each mode builds on the previous:

  • problem → frames who, what, and why
  • competitive → validates market gap
  • feasibility → validates buildability
  • checkpoint → synthesizes all three into a go/no-go verdict

You can run individual modes out of order when the user has already done partial research.

Step 4: Checkpoint Verdict

After the checkpoint mode:

  • Go → "Run /story create to define the feature scope."
  • Conditional go → Flag caveats, suggest mitigations, then allow forward.
  • No-go → Explain which criteria failed, suggest what needs to change before re-evaluating.

Template Locations

Summary Templates (~20 lines each — L1 default)

| Mode | File | | ------------- | ------------------------------------------------------------- | | overview | ~/.claude/skills/discovery/templates/overview.md | | problem | ~/.claude/skills/discovery/templates/problem/summary.md | | competitive | ~/.claude/skills/discovery/templates/competitive/summary.md | | feasibility | ~/.claude/skills/discovery/templates/feasibility/summary.md | | checkpoint | ~/.claude/skills/discovery/templates/checkpoint/summary.md |

Patterns Templates (~60 lines each — L2 on request)

| Mode | File | | ------------- | -------------------------------------------------------------- | | problem | ~/.claude/skills/discovery/templates/problem/patterns.md | | competitive | ~/.claude/skills/discovery/templates/competitive/patterns.md | | feasibility | ~/.claude/skills/discovery/templates/feasibility/patterns.md | | checkpoint | ~/.claude/skills/discovery/templates/checkpoint/patterns.md |

Progressive Disclosure Rules

  1. Always start at L1 — never present all 4 modes' full patterns at once
  2. Read templates on demand — use the Read tool; don't memorize template content
  3. User drives depth — "show me how to score", "give me the rubric", "examples?" triggers L2
  4. Don't hard-block the user — if they skip modes, accept partial input but flag explicitly which criteria are scored on assumptions rather than evidence
  5. Checkpoint is the exit gate — nothing moves to /story create without a checkpoint verdict

Integration

Position in pipeline: Phase 0 → before everything else

/discovery (Phase 0)
  → /story create (Phase 1 — only after checkpoint go)
    → /design (Phase 2)
      → /implementation (Phase 3)

Invoked by: User directly (/discovery) Invokes: Nothing — produces validated problem definition consumed by /story create Connected skills:

  • /story create — next step after checkpoint go
  • /design — invoked by user after /story prp generates a PRP; uses the validated problem scope from discovery
  • /product-assess — deeper product assessment for complex ideas (optional, before /discovery checkpoint)

Examples

Full sequence on a new idea:

User: /discovery problem → frames who/urgency/evidence User: /discovery competitive → maps alternatives User: /discovery feasibility → checks buildability User: /discovery checkpoint → verdict: Go → "Great, run /story create to define the feature."

Skipping straight to checkpoint:

User: /discovery checkpoint → Present checkpoint summary, fill criteria based on what user already knows → Flag any missing evidence as risks

L2 escalation:

User: /discovery problem → Show problem/summary.md User: "show me how to score urgency" → Read and show problem/patterns.md (pain-point scoring section)