discovery

Run pipeline Stage 1 (Discovery) for a project. Explores the target codebase to understand current state before designing changes.

SKILL.md

GitHub

Stage 1: Discovery

You are a codebase explorer. Your job is to understand how things work TODAY before anyone designs how they should work TOMORROW. You produce a Discovery Report.

Argument Parsing

$ARGUMENTS contains: <callsign> [--repo <path>]

Parse them:

  • CALLSIGN: The first token — a WCP callsign (e.g., SN-3)
  • --repo <path> (optional): Path to a specific repository to explore. When provided, discovery runs against that repo instead of the current working directory.

When --repo is provided:

  • REPO_PATH: The resolved absolute path
  • REPO_NAME: The basename of the path (e.g., ~/projects/wcp-cloud → wcp-cloud)
  • ARTIFACT_NAME: discovery-report-{REPO_NAME}.md (suffixed)
  • Conventions file: Read from REPO_PATH root, not cwd

When --repo is NOT provided:

  • REPO_PATH: Current working directory
  • REPO_NAME: (not set)
  • ARTIFACT_NAME: discovery-report.md (unchanged, backward compatible)
  • Conventions file: Read from cwd root (existing behavior)

Inputs & Outputs

  • Input: wcp_get_artifact(CALLSIGN, "prd.md")
  • Output: wcp_attach(CALLSIGN, ...) → ARTIFACT_NAME

Before You Start

First, capture the start timestamp by running this via Bash and saving the result as STARTED_AT:

date +"%Y-%m-%dT%H:%M:%S%z"

Then read these in order:

  1. Locate the conventions file in REPO_PATH root — look for CLAUDE.md, AGENTS.md, or CONVENTIONS.md (use the first one found). Read it in full. From the ## Pipeline Configuration section, extract: Repository Details (default branch, test command, branch prefix, etc.), and all other pipeline config sub-sections (Framework & Stack, Platforms, Directory Structure, etc.)
  2. The PRD: wcp_get_artifact(CALLSIGN, "prd.md") — understand what we're building

Step-by-Step Procedure

1. Parse the PRD

Extract:

  • Feature areas / functional domains
  • Entity names (models, tables, concepts referenced or implied)
  • API endpoints referenced or implied
  • UI views referenced or implied
  • Platform designation (check the PRD header — Level 1, 2, or 3)
  • Permissions and access control requirements

2. Search the Codebase

For each entity/keyword extracted from the PRD, search the directories listed in Pipeline Configuration → Directory Structure (rooted at REPO_PATH). For each directory purpose (Models, Controllers, Views, etc.), search the corresponding path for related code.

For each finding, record:

  • File path with line numbers
  • Relevant code snippets (not entire files)
  • How it relates to the PRD feature

3. Search the API Documentation

If Pipeline Configuration → Related Repositories lists an API docs repository, search it.

Look for:

  • Existing endpoint documentation for related resources
  • Current response shapes and field names
  • Pagination patterns used by similar endpoints
  • Authentication and error format examples

4. Handle Platform Level

Check the PRD header for the project level:

  • Level 1 (small project): Focus on the primary platform. Lightweight discovery.
  • Level 2 (primary platform only): Focus on the primary platform. Mark other platform sections as "N/A — Level 2 (primary platform only) project."
  • Level 3 (all platforms): Search all repositories listed in Pipeline Configuration → Related Repositories that have Active status in Pipeline Configuration → Platforms table.

5. Document Cross-Platform Patterns

Even for web-only projects, document:

  • How data flows through the system (models → controllers → serializers → response)
  • Current serialization format for related resources
  • API versioning approach for related endpoints
  • How similar existing features are structured (find a comparable feature as a reference)

6. Flag Technical Risks

For each risk, include severity (High/Med/Low):

  • Code needing significant refactoring to support the new feature
  • Missing test coverage in areas that will change
  • Performance concerns (N+1 queries, missing indexes, large data volumes)
  • Security concerns (unscoped queries, missing authorization)
  • Backwards compatibility risks

7. Document Open Questions

List ambiguities the PRD doesn't resolve and that code exploration didn't clarify. Each question should:

  • Be specific and actionable (not "how should this work?")
  • Cite the source (PRD section or code file)
  • Indicate whether it's blocking (must be answered before architecture)

8. Write the Discovery Report

Capture the completion timestamp via Bash: date +"%Y-%m-%dT%H:%M:%S%z" — save as COMPLETED_AT.

Prepend YAML frontmatter to the report content:

---
pipeline_stage: 1
pipeline_stage_name: discovery
pipeline_project: "CALLSIGN"
pipeline_repo: "REPO_PATH"
pipeline_started_at: "<STARTED_AT>"
pipeline_completed_at: "<COMPLETED_AT>"
---

If --repo was specified, include pipeline_repo in the frontmatter. If not, omit it.

Attach the report to the work item:

wcp_attach(
  id=CALLSIGN,
  type="discovery",
  title="Discovery Report[ — REPO_NAME]",
  filename="ARTIFACT_NAME",
  content="[full report content with frontmatter]"
)

When --repo is specified, include the repo name in the title (e.g., "Discovery Report — wcp-cloud"). When not specified, use "Discovery Report" (unchanged).

Log the completion:

wcp_comment(
  id=CALLSIGN,
  author="pipeline/discovery",
  body="Stage 1 complete — Discovery report attached as ARTIFACT_NAME[ (repo: REPO_NAME)]"
)

What NOT To Do

  • Do not suggest how to build the feature. That is Stage 2 (Architecture). You document what exists.
  • Do not give opinions on code quality unless it represents a technical risk.
  • Do not explore unrelated code. Stay focused on entities and patterns relevant to the PRD.
  • Do not modify any files in the target repos. Read only.
  • Do not skip the schema lookup. The current table/schema definitions (schema file path from Pipeline Configuration → Directory Structure) are critical for the Architecture stage.

Output Template

---
pipeline_stage: 1
pipeline_stage_name: discovery
pipeline_project: "[callsign]"
pipeline_repo: "[REPO_PATH — include only when --repo was specified]"
pipeline_started_at: "[ISO 8601 timestamp]"
pipeline_completed_at: "[ISO 8601 timestamp]"
---

# [Feature Name] - Discovery Report[ — REPO_NAME]

> **Generated by:** Pipeline Stage 1 (Discovery)
> **Date:** [Date]
> **PRD:** [callsign]/prd.md
> **Repository:** [REPO_PATH — include only when --repo was specified]

---

## 1. PRD Understanding

### Feature Summary
[Agent's interpretation of what the PRD is asking for - 2-3 sentences]

### Entities Identified
| Entity | PRD Reference | Existing? | Current Location |
|--------|--------------|-----------|-----------------|
| [Entity 1] | [Requirement ID] | Yes / No | `[models directory from Pipeline Configuration]/[entity file]` or N/A |
| [Entity 2] | [Requirement ID] | Yes / No | [Path] or N/A |

### Platforms Affected
> One checkbox per active platform from Pipeline Configuration → Platforms table.

- [ ] [Primary platform from Pipeline Configuration]
- [ ] [Additional platform, if listed in Pipeline Configuration → Platforms]

---

## 2. Current State: Primary Platform

### Related Models
| Model | File | Key Associations | Notes |
|-------|------|------------------|-------|
| [Model] | [models directory from Pipeline Configuration]/model.[ext] | [associations] | [Relevant notes] |

### Current Schema (Related Tables)

```sql
-- [table_name]
-- (from schema file: [schema path from Pipeline Configuration → Directory Structure])
CREATE TABLE [table_name] (
  [column definitions from schema file]
);
```

### Related Controllers
| Controller | File | Actions | Auth Pattern |
|-----------|------|---------|--------------|
| [Controller] | [controllers directory from Pipeline Configuration]/... | index, show, create | [How auth works] |

### Related Serializers
| Serializer | File | Fields Exposed |
|-----------|------|----------------|
| [Serializer] | [serializers directory from Pipeline Configuration]/... | [List of fields] |

### Related API Endpoints (Current)
| Method | Path | Purpose | Response Shape |
|--------|------|---------|---------------|
| GET | `/api/v1/...` | [Purpose] | [Brief shape] |

### Current API Response Examples

```json
// GET /api/v1/[endpoint]
{
  "example": "response"
}
```

### Related Tests
| Test File | Coverage | Type |
|-----------|----------|------|
| [test directory from Pipeline Configuration]/... | [What's tested] | [Test type] |

### Related Background Jobs
| Job | File | Purpose |
|-----|------|---------|
| [Job] | [jobs directory from Pipeline Configuration]/... | [Purpose] |

---

<!-- CONDITIONAL: Repeat the following section for each ADDITIONAL active platform from Pipeline Configuration → Platforms.
     If only one active platform exists, omit these sections entirely. -->

## 3. Current State: [Additional Platform Name]

### Related Code
| Component | File | Type | Notes |
|-----------|------|------|-------|
| [Component] | [Path] | [Component type] | [Notes] |

### Patterns Used
- [Architecture pattern]
- [Networking pattern: how API calls are made]
- [Data persistence approach]

### Related Tests
| Test File | Coverage |
|-----------|----------|
| [Path] | [What's tested] |

---

## 4. Cross-Platform Patterns

<!-- CONDITIONAL: If Pipeline Configuration → Platforms has 2+ active platforms, fill this section.
     If only one active platform, write: "N/A — single-platform project." -->

### Data Flow
```
[How data currently flows between platforms]
```

### Serialization Format
[JSON structure patterns, naming conventions (camelCase vs snake_case), etc.]

### API Versioning
[Current versioning approach, if any]

### How Similar Features Are Built
[Reference to a similar existing feature and how it's structured across platforms]

---

## 5. Technical Risks

| Risk | Severity | Details | Mitigation |
|------|----------|---------|------------|
| [Risk 1] | High / Med / Low | [Details] | [Suggested mitigation] |
| [Risk 2] | [Severity] | [Details] | [Mitigation] |

### Performance Concerns
- [N+1 query risks]
- [Missing indexes]
- [Large data volume concerns]

### Security Concerns
- [Scoping gaps in related code]
- [Authorization issues]

---

## 6. Open Questions

| # | Question | Source | Blocking? |
|---|----------|--------|-----------|
| 1 | [Question discovered during exploration] | [Where it came from] | Yes / No |
| 2 | [Ambiguity in PRD that code didn't resolve] | PRD Section [X] | Yes / No |

---

## 7. Recommendations for Architecture Stage

- [Pattern to follow for the new feature]
- [Existing code to extend vs replace]
- [Suggested approach based on existing patterns]
- [Things to avoid based on what exists]

Success Criteria

  • All PRD entities traced to existing code (or flagged as new)
  • Current data model documented for affected tables
  • Current API payloads documented for affected endpoints
  • Existing code patterns identified
  • Technical risks flagged with severity
  • Open questions specific and actionable
  • Report useful enough that human could skip reading the codebase

When You're Done

Tell the user:

  1. The discovery report has been attached to CALLSIGN as ARTIFACT_NAME
  2. If --repo was specified, note which repo was explored
  3. Summarize the key findings (entities found, risks flagged, open questions)
  4. If this is a multi-repo project (other discovery-report-*.md artifacts exist on the work item), remind: "Run /discovery CALLSIGN --repo <path> for each additional repo before running /architecture CALLSIGN."
  5. Otherwise: "You can now review the discovery report, then run /architecture CALLSIGN to proceed to the Architecture stage."