# Authentication

> Authentication

- Skill: `daboggieman/authentication` (Agent Skill)
- Install (CLI): `npx skillmds@latest add daboggieman/authentication`
- Raw SKILL.md: https://api.skillmd.com/api/skills/daboggieman/authentication/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: Daboggieman (https://skillmd.com/u/daboggieman)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/daboggieman/authentication

---

# Authentication

## Purpose

Implement and review user identity, login, registration, sessions, credentials, and authentication flows.

## Project Stack

FastAPI, JWT, bcrypt, Next.js.

## Operating Principles

1. Never store plaintext passwords.
2. Use the project's established password hashing implementation.
3. Keep secrets in environment configuration.
4. Validate credentials on the server.
5. Do not expose sensitive authentication information in API responses.
6. Keep token expiration and refresh behavior explicit.
7. Protect authentication endpoints against common abuse patterns.
8. Do not log passwords, tokens, reset links, or sensitive credentials.
9. Treat client-side authentication state as untrusted.
10. Verify authentication behavior with actual requests where possible.

## Required Workflow

1. Inspect existing authentication implementation.
2. Trace registration/login/token/session flow.
3. Identify trust boundaries.
4. Research current framework security behavior if uncertain.
5. Implement the smallest safe change.
6. Test valid and invalid authentication cases.
7. Test expiration and unauthorized access.
8. Review logging and secret handling.
9. Run security review.

## Evidence Rules

Use the strongest available evidence in this order:

1. Runtime behavior
2. Automated test results
3. Type checking, linting, and build results
4. Installed package/source behavior
5. Official framework or library documentation
6. Repository configuration
7. Static code inspection
8. General model knowledge

Never report an assumption as a verified result.

## Tool Selection

- Use Filesystem MCP for repository exploration and file inspection.
- Use Context7 for current, version-specific library/framework documentation.
- Use Fetch when official documentation or public web resources need to be retrieved.
- Use Playwright for browser behavior, frontend interaction, forms, authentication flows, and UI verification.
- Use GitHub CLI for repository, branch, pull request, issue, workflow, and CI operations.
- Prefer the smallest tool set that establishes reliable evidence.

## Failure Handling

When something fails:

1. Reproduce the failure.
2. Capture the exact error.
3. Identify the failing layer.
4. Check whether the failure is environmental, dependency-related, configuration-related, or caused by application code.
5. Research uncertain behavior using authoritative sources.
6. Apply the smallest appropriate fix.
7. Reproduce the original failure again.
8. Run regression checks.
9. Inspect the final diff.

Do not hide failures by weakening tests, suppressing errors, deleting evidence, or claiming success without verification.

## Security

- Never expose secrets.
- Never commit credentials.
- Treat all external input as untrusted.
- Preserve authentication and authorization boundaries.
- Avoid leaking internal implementation details through errors.
- Review security implications whenever the skill touches users, data, credentials, networking, or deployment.

## Completion Criteria

A task using this skill is complete only when:

- The requested behavior is implemented.
- Existing project conventions are preserved.
- Relevant static checks pass.
- Relevant tests pass or their absence is explicitly reported.
- Browser verification is performed when the task affects user-facing behavior.
- Security implications have been considered.
- The final diff has been inspected.
- Verified facts are clearly distinguished from assumptions or remaining uncertainty.