Symmetric Dogfooding

Bidirectional integration validation pattern where two repositories each consume the other for testing, ensuring both sides work correctly together before downstream adoption.

Self-Evolving Skill: This skill improves through use. If instructions are wrong, parameters drifted, or a workaround was needed — fix this file immediately, don't defer. Only update for real, reproducible issues.

Pattern Overview

┌─────────────────────────────────────────────────────────────────┐
│                    SYMMETRIC DOGFOODING                         │
│                                                                 │
│        Repo A ◄─────── mutual validation ───────► Repo B        │
│                                                                 │
│   EXPORTS:                              EXPORTS:                │
│   - Library/API                         - Library/API           │
│   - Data structures                     - Data structures       │
│                                                                 │
│   VALIDATES WITH:                       VALIDATES WITH:         │
│   - Repo B real outputs                 - Repo A real outputs   │
│   - Production-like data                - Production-like data  │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘

When to Use This Skill

Use this skill when:

• Two repos have a producer/consumer relationship
• APIs evolve independently and need integration testing
• Data formats may drift between repos
• Both repos are actively developed

---

TodoWrite Task Templates

Template A: Setup Symmetric Dogfooding Between Two Repos

1. Identify integration surface (exports from A consumed by B and vice versa)
2. Document data formats, schemas, API signatures at boundary
3. Configure cross-repo dev dependencies in both repos
4. Pin versions explicitly (tags or SHAs, never main)
5. Create integration/ test directory in both repos
6. Write bidirectional validation tests (A validates with B outputs, B validates with A outputs)
7. Add validation tasks to mise.toml or Makefile
8. Document pre-release protocol in both CLAUDE.md files
9. Run full symmetric validation to verify setup
10. Verify against Symmetric Dogfooding Checklist below

Template B: Pre-Release Validation

1. Run validate:symmetric task in releasing repo
2. Check if other repo has pending changes affecting integration
3. If yes, test against other repo's feature branch
4. Document any failures in validation log
5. Fix integration issues before release
6. Update version pins after successful validation
7. Coordinate if breaking changes require simultaneous release
8. Verify against Symmetric Dogfooding Checklist below

Template C: Add New Integration Point

1. Identify new export/import being added
2. Update integration surface documentation
3. Add tests in both repos for new integration point
4. Run symmetric validation in both directions
5. Update version pins if needed
6. Verify against Symmetric Dogfooding Checklist below

Symmetric Dogfooding Checklist

After ANY symmetric dogfooding work, verify:

• Both repos have integration tests that import the other
• Version pins are explicit (tags or commit SHAs)
• Pre-release checklist includes cross-repo validation
• Integration tests use real data (not mocks of the other repo)
• Breaking changes coordination documented
• Validation task runnable via single command

---

Post-Change Checklist (Self-Maintenance)

After modifying THIS skill:

1. Templates cover common symmetric dogfooding scenarios
2. Checklist reflects current best practices
3. Example in references/ still accurate
4. Append changes to evolution-log.md

---

Implementation Guide

Phase 1: Discovery and Mapping

Identify the integration surface:

• List all exports from Repo A consumed by Repo B
• List all exports from Repo B consumed by Repo A
• Document data formats, schemas, API signatures

Map validation scenarios:

• What real-world data from B can validate A outputs?
• What real-world data from A can validate B outputs?
• Identify edge cases that only appear in production usage

Phase 2: Dependency Configuration

Configure cross-repo dev dependencies:

Python (uv/pip):

# Repo A pyproject.toml
[project.optional-dependencies]
validation = ["repo-b"]

[tool.uv.sources]
repo-b = { git = "https://github.com/org/repo-b", tag = "<tag>" }  # SSoT-OK

# Repo B pyproject.toml
[project.optional-dependencies]
validation = ["repo-a"]

[tool.uv.sources]
repo-a = { git = "https://github.com/org/repo-a", tag = "<tag>" }  # SSoT-OK

Rust (Cargo):

[dev-dependencies]
repo-b = { git = "https://github.com/org/repo-b", tag = "<tag>" }  # SSoT-OK

Node.js:

{
  "devDependencies": {
    "repo-b": "github:org/repo-b#<tag>"
  }
}

Critical: Pin to tags or commit SHAs. Never use main/master branches.

Phase 3: Test Infrastructure

Directory structure in both repos:

repo-a/
└── tests/
    ├── unit/              # Internal tests
    └── integration/       # Tests using repo-b real outputs
        └── test_with_repo_b.py

repo-b/
└── tests/
    ├── unit/              # Internal tests
    └── integration/       # Tests using repo-a real outputs
        └── test_with_repo_a.py

Bidirectional validation test pattern:

# repo-a/tests/integration/test_with_repo_b.py
"""Validate Repo A outputs work correctly with Repo B inputs."""

def test_a_output_consumed_by_b():
    # Generate output using Repo A
    a_output = repo_a.generate_data()

    # Feed to Repo B - should work without errors
    b_result = repo_b.process(a_output)

    # Validate the round-trip
    assert b_result.is_valid()

Phase 4: Task Automation

mise.toml example:

[tasks."validate:symmetric"]
description = "Validate against partner repo"
run = """
uv sync --extra validation
uv run pytest tests/integration/ -v
"""

[tasks."validate:pre-release"]
description = "Full validation before release"
depends = ["test:unit", "validate:symmetric"]

Phase 5: Pre-Release Protocol

Before releasing Repo A:

1. Run validate:symmetric in Repo A (tests against current Repo B)
2. If Repo B has pending changes, test against Repo B branch too
3. Update version pins after successful validation

Before releasing Repo B:

1. Run validate:symmetric in Repo B (tests against current Repo A)
2. If Repo A has pending changes, test against Repo A branch too
3. Update version pins after successful validation

Coordinating breaking changes:

• If A needs to break compatibility, update B first
• If B needs to break compatibility, update A first
• Consider simultaneous releases for tightly coupled changes

---

Anti-Patterns

Anti-Pattern	Problem	Solution
One-direction only	Misses half the bugs	Always test both directions
Using main branch	Unstable, breaks randomly	Pin to tags or SHAs
Skipping for small changes	Small changes cause big breaks	Always run full validation
Mocking partner repo	Defeats the purpose	Use real imports
Ignoring version matrix	Silent production failures	Maintain compatibility matrix

---

References

• example-setup.md - Real-world trading-fitness/rangebar-py example
• evolution-log.md - Skill change history

External:

• Dogfooding (DevIQ)
• CDC Testing (Microsoft)

---

Troubleshooting

Issue	Cause	Solution
Dependency resolution fails	Version pin outdated	Update tag/SHA pin to latest stable version
Tests pass locally fail CI	Different partner repo version	Pin exact same version in both environments
Breaking change not caught	One-direction testing only	Run validate:symmetric in BOTH repos
Integration surface unclear	Undocumented exports	Map all imports/exports before setting up tests
Too many parts moving	Uncoordinated releases	Coordinate breaking changes, test branches first
Mock data hiding bugs	Using stubs instead of real	Always import real partner repo for integration
Version matrix explosion	Too many combinations	Limit support to N-1 versions, document clearly
Circular dependency	Both repos require each other	Use optional-dependencies for validation only

Post-Execution Reflection

After this skill completes, reflect before closing the task:

0. Locate yourself. — Find this SKILL.md's canonical path before editing.
1. What failed? — Fix the instruction that caused it.
2. What worked better than expected? — Promote to recommended practice.
3. What drifted? — Fix any script, reference, or dependency that no longer matches reality.
4. Log it. — Evolution-log entry with trigger, fix, and evidence.

Do NOT defer. The next invocation inherits whatever you leave behind.