Claude Skills by BrennonTWilliams

Use when asked for a second opinion from a different, stronger model on a specific decision, question, or stalled judgment call.

Use when asked to detect conflicting requirements or incompatible decisions across open issues.

Use when asked to capture or create an issue from conversation or natural language.

Use when asked for a pre-implementation confidence check or whether an issue is ready to implement.

Use when asked to configure little-loops, change settings, or update ll-config.json.

Use when asked to generate an FSM eval harness YAML from one or more issue IDs, or DSL eval tasks from a loop/issue file with --dsl.

Use when asked to create an automation loop or automate a recurring task with an FSM.

Use when asked to select the winning implementation option for an issue with decision_needed.

Use when exploring an external API/library and recording proof to the Learning Test Registry.

Use when asked to format an issue or fix its template structure.

Use when asked for an adversarial go/no-go review or whether an issue is worth implementing.

Use when asked to implement an issue end-to-end or start implementing FEAT-NNN.

Use when asked to analyze product goals, check feature gaps, or evaluate business value. Returns raw YAML findings. For full scan with issue file creation, use `/ll:scan-product`.

Use when asked to review epic health, audit stalled children, check scope drift, find missing coverage, or assess closure readiness. Produces a structured health report and actionable recommendations. Read-only.

Use when asked to decompose a theme or goal into an EPIC with 3–8 child issues. Creates the EPIC file, pre-wired child stubs, and stages everything for git.

Use when asked to prove an unproven internal mechanism with an isolated code spike before implementing.

Use when asked to generate an FSM verification loop YAML from a single issue's acceptance criteria, or an adversarial variant that tries to break the feature via boundary/malformed/failure-mode probes.

Use when a refined issue is missing integration points or wiring in the implementation plan.

Use when asked to propose automations or suggest automation opportunities from workflow patterns.

Create a new skill `/ll:issue-size-review` that evaluates the size/complexity of active issues, identifies those unlikely to be completed in a single session, and proposes decomposition into multiple smaller issues with user approval.

The `/ll:configure` skill (`skills/configure/`) has no awareness of `loops.run_defaults`. The setting added in ENH-2109 (`clear`, `show_diagrams`, `mode` under `loops.run_defaults`) cannot be discovered or set via the interactive configure flow — users must hand-edit `ll-config.json` directly.

`ll-logs scan-failures --project <path>` reports failure clusters keyed by tool+error signature across the entire project, with no way to filter to a single skill (e.g. `review-epic`). Getting skill-specific analytics currently requires running the full scan and manually grepping the output for the skill name.

Update `/ll:confidence-check` to support `--all` and `--auto` flags, matching the batch-processing pattern established by `/ll:ready-issue`. Additionally, ensure the skill updates a `confidence_score` field in the issue's YAML frontmatter after evaluation.

The `/ll:commit` skill should accept an `--auto` flag that suppresses all user interactivity and makes the best decision it can autonomously — staging files, writing commit messages, and executing without prompting.

Add a **Step 6: Config Health Check** to `skills/update/SKILL.md` that runs after a successful update and validates `.ll/ll-config.json` against `config-schema.json`. Reports unknown/additional keys and type mismatches; prints `[PASS] ll-config.json is valid` or `[WARN] Config issues detected`.

Add an interactive `/ll:review-sprint` skill that uses Claude's intelligence to analyze a sprint and suggest improvements — reordering waves for efficiency, swapping issues based on priority changes, recommending issues to remove or add from the backlog, and resolving dependency conflicts. This goes beyond the deterministic edits in `ll-sprint edit` (ENH-393) by providing opinionated, context-aware recommendations.

Create a `product-analyzer` **skill** that analyzes the codebase against product goals to identify: - Feature gaps (goals without implementation) - User experience improvements - Business value opportunities - Strategic alignment issues

Create `workflow-automation-proposer` **skill** (Step 3 of the workflow analysis pipeline) that synthesizes patterns and workflows into concrete automation proposals including slash commands, scripts, and hooks.

Add a `.ll/learning-tests/` registry directory that stores executed proof records of external system behaviors, paired with a new `ll:explore-api` skill that implements the full Feathers Learning Test lifecycle (Ingest → Hypothesize → Execute → Refine). Together they give agents institutional knowledge of proven assumptions before committing to architecture.

Implement `skills/explore-api/SKILL.md` — a new skill that guides agents through the full Feathers Learning Test lifecycle (Ingest → Hypothesize → Execute → Refine) for any external system target, then writes a proof record to the `.ll/learning-tests/` registry via `ll-learning-tests`. Also includes the count/table updates to README, CONTRIBUTING, CLAUDE.md, and ARCHITECTURE.md that completing the full feature requires.

little-loops had a family of loop-operating skills (`create-loop`, `review-loop`, `rename-loop`, `cleanup-loops`, `debug-loop-run`, `audit-loop-run`) but none that *refactors* a loop's structure. Loops accrete: a single `states:` map grows to dozens of states, long linear runs are written verbosely, and self-contained phases stay inline instead of becoming reusable sub-FSMs. This adds `/ll:simplify-loop <name>` — a skill that takes a project-level or built-in FSM loop and applies two **behavior-

Create `skills/audit-issue-conflicts/SKILL.md` — the primary skill file implementing conflict detection, recommendation synthesis, interactive approval loop, and `--auto`/`--dry-run` modes.

Add a `/ll:review-epic <EPIC-ID>` skill that audits an EPIC's children for (1) stalled status (open/blocked for N days with no activity), (2) scope drift (child summaries that no longer match the EPIC's stated scope), (3) missing coverage (EPIC scope mentions areas with no child issue), and (4) closure readiness (all children done → recommend marking EPIC done). Output is a structured health report plus actionable suggestions, mirroring the shape of `/ll:review-sprint`.

Add a `/ll:scope-epic` skill that takes a high-level theme/goal description and produces (1) an EPIC issue file scoped to the theme and (2) 3–8 child issue stubs pre-wired with `parent: EPIC-NNN`. This is the upstream creation step that `/ll:capture-issue --parent` assumes already happened.

Create a new `/ll:` Agent Skill that uses `ll-messages` output to analyze user message history and suggest new FSM loops with 3-15 steps, bypassing the interactive `/ll:create-loop` wizard.

Create a new `/ll:refine-issue` skill that accepts an Issue ID (e.g., FEAT-XXX), reads the issue file, and asks the user clarifying questions to refine the issue before implementation.

Add a new `/ll:create-epics-from-unparented` skill (or an `--create` flag on `/ll:link-epics`) that clusters orphaned issues by thematic similarity and proposes new EPIC definitions to cover them — the inverse operation of `/ll:link-epics`. The current `/ll:link-epics` skill only links unparented issues to **existing** EPICs; when no EPICs exist yet (or when orphaned issues don't fit any existing EPIC), users have no guided way to synthesize EPICs from the orphan pool.

Extend `/ll:align-issues` to support aligning issues against a specific skill (e.g., `/ll:align-issues skill capture-issue`). Currently the command aligns issues against key documents in configured categories. This feature adds a parallel path that uses a skill's `SKILL.md` as the alignment reference, so users can check whether an issue is well-scoped for a particular skill to implement.

Validate active issues against key documents for relevance and alignment

Analyze user message history to identify patterns, workflows, and automation opportunities

Analyze codebase architecture for patterns, organization, and improvements

Run code quality checks (lint, format, types, build)

Clean orphaned git worktrees from interrupted ll-parallel/ll-loop runs

Create git commits with user approval and no assistant attribution

Create a sprint definition with a curated list of issues

Generate comprehensive PR descriptions following repository templates

Analyze codebase for deprecated, unused, or dead code that can be safely removed

Generate continuation prompt for session handoff

List all available little-loops commands with descriptions

Iterate on existing implementation plans with thorough research and updates