lexedwards/daedalus

Review and improve professional job-application documents such as CVs, resumes, and cover letters. Use only when those materials are supplied or requested; do not use for personal websites, portfolios, project pages, biographies, social profiles, or general professional branding.

Guide pragmatic implementation, bug fixing, refactoring, and review with isolated work, verified increments, and PR delivery by default, using minimal design and risk-based verification.

Create agent skills. Use when authoring a new skill or asking about SKILL.md

Capture important and hard-to-reverse decisions in ADRs as they crystalise. Use when a user wants to preserve important or critical implementation decisions.

Extract clean markdown content from web pages using Defuddle CLI - removing clutter and navigation. Use instead of webfetch or WebFetch if the url **does not** end in .md **and** the user wants to read online web pages, articles, documentation.

Evaluate a cited change to an agent environment through realistic, blinded before-and-after work informed by the current project. Use only when the user explicitly invokes eval or requests this skill.

Author, compare, and troubleshoot custom skill and agent configuration across OpenCode V2, Codex, Claude Code, Cursor, GitHub Copilot CLI, and Pi. Use when checking supported fields, discovery, precedence, portability, or updating this skill's own configuration references.

Edit prose to preserve the writer's voice while checking surface AI tells and predictable choices in structure, explanation, and storytelling. Use for essays, articles, profiles, case studies, fiction, and other AI-sounding writing.

Interview and challenge the user until reaching a shared understanding, resolving each branch of the decision tree. If present, contrast against existing domain model, sharpening terminology and updating documentation as decisions crystallise. Use when a user wants to discuss an idea or stress-test a plan with or without existing project documentation.

Plan incremental implementation by breaking a plan, spec, PRD, Jira Epic, or Linear project or project milestone into vertical work slices. Use when the user wants to create implementation tasks or break down work into manageable tasks.

Checkpoint implementation through small verified commits using Conventional Commits, meaningful component scopes, and references to known issues in the task context. Use during implementation or when preparing commits, whether or not the work has a tracked issue.

Initialize, maintain, or reference a LLM Wiki vault. A personal knowledge base and AI second brain in a specific Karpathy-style structure. Complete with AGENTS.md, scaffolded directory structure and indexing.

Publish verified implementation slices as pull requests, including dependent stacks, and handle requested CI and review follow-up. Use when delivering code, opening or updating PRs, or preparing work for integration.

Apply when placing validation, error handling, domain logic, or framework adapters. Validate at trust boundaries, use domain values internally, and keep the I/O shell thin.

Apply when investigating, reviewing, or reporting conclusions. Support verdicts with inspected evidence and distinguish defects from uncertainty and preferences.

Apply when designing modules, interfaces, or abstractions. Hide cohesive complexity behind a small interface that reduces caller knowledge.

Apply when delivering implementation work or deciding whether it is complete. Integrate changes through PRs with central checks and independent review rather than equating local verification with delivery.

Apply before writing or delegating work when actors may share files, branches, runtime resources, or other mutable state. Separate independent writers before coordinating unavoidable shared writes.

Apply when making consequential decisions or documenting changed behavior. Preserve non-obvious rationale, constraints, and rejected alternatives for future maintainers.

Apply when planning or executing multi-step work, migrations, or delivery. Build complete, useful slices and verify each before building on it.

Apply when sizing a change, fixing a defect, or considering new code. Make the smallest coherent change that satisfies the outcome across affected surfaces.

Apply when writing, changing, or assessing tests. Exercise public behavior and assert observable outcomes against independent expected values.

Apply before changing code, diagnosing a defect, or assessing a design. Ground decisions in the actual behavior, callers, and boundaries.

Apply when choosing checks, review effort, or completion evidence. Match verification to affected boundaries, consequences, and observed failures.

Capture product requirements for new or changed user-facing work before implementation. Use when the user wants to capture requirements, create or update a local .specs/<spec-name>/REQUIREMENTS.md spec, Jira Epic, Linear project or project milestone, or draft/update a PRD-style product requirements document in the response unless a destination is specified.

Help the user understand the current topic visually with concise diagrams, code-shape sketches, and focused artifacts.

Write technical documentation that is approachable and concise. Use when recording context that future engineers and agents will need to understand and operate.