Skills

All plugin functionality is delivered through skills. The Invocation column shows who can start each one:

• User: only you, from the host's / menu ($ in Codex). These skills set disable-model-invocation: true and never auto-run.
• User, Auto: you, or the host when the context matches the skill.
• Background: only the host, automatically. These skills set user-invocable: false and don't appear in the menu.

Each row gives a one-line summary. The full description lives in the skill's SKILL.md, and the generated component catalog indexes every skill, agent, and hook by name, invocation mode, and source path.

Find a Skill by Task

I want to…	Use
Find which trackbed skill fits my task, or what to run next	/trackbed:workflow:suggest-next
Turn a vague idea into clear requirements	/trackbed:discovery:interview
Chart an initiative too big or foggy to plan in one session	/trackbed:discovery:wayfinder
Get a cited answer from primary sources	/trackbed:research
Write a feature spec	/trackbed:docs:feature-spec, or /trackbed:siw:init for tracked work
Write a PRD for the prototype I'm shipping	/trackbed:product:prd
Record an architecture decision	/trackbed:docs:adr
Create or refine a Linear issue	/trackbed:linear:issue-define
Pick the next Linear issue to work on	/trackbed:linear:next
Implement and ship a Linear issue	/trackbed:linear:issue-to-pr
Implement a saved plan	/trackbed:code:plan-to-pr or /trackbed:code:work-from-plan
Review my branch	/trackbed:pr:code-review, or see which review to use
Open a PR	/trackbed:pr:create
Fix failing CI on my PR	/trackbed:pr:fix-ci
Get my open PR ready to merge: rebase, green CI, and AI reviewer comments answered	/trackbed:pr:shape-up
See which PRs wait on my review	/trackbed:pr:review-queue
Review someone else's PR	/trackbed:pr:github-review
Respond to review comments on my PR	/trackbed:pr:github-review-reply
Run tests, linting, and type checks	/trackbed:verify:run
Investigate a bug	/trackbed:debug:investigate
Test a running app	/trackbed:qa
Find refactoring candidates or tech debt	/trackbed:code:refactor-opportunities, or add --for <spec> before building a planned change
Stop agents from repeating a mistake by turning it into a type, lint rule, or check	/trackbed:code:encode-constraint
Upgrade a framework or library	/trackbed:code:migrate
Explain a system or change visually	/trackbed:visual:explain
Make a document easier to read	/trackbed:text:clarify

User-Invocable Skills

Structured Implementation Workflow (SIW)

Local specification refinement and initial issue planning before a one-way Linear transfer. See docs/siw.md.

Skill	Invocation	Arguments	Description
/trackbed:siw:init	User	[spec-file(s) | folder | discover] [--auto]	Initialize SIW documents in siw/: the spec, LOG.md, and issues.
/trackbed:siw:generate-phases	User	[spec-file-path] [--auto]	Break the spec into atomic, phase-based issues with tests and validation.
/trackbed:siw:spec-audit	User	[spec-file-path(s) | 'siw' | audit-report-path] [--product] [--auto] [--apply [--dry-run] [--threshold 60-100] [--allow-dirty]] [--model opus|sonnet|haiku] [--inline] [--team]	Audit specs for implementation readiness; --product adds product checks and --apply fixes mechanical findings.
/trackbed:siw:resolve-audit	User	[audit-report-path] [finding-id(s)] [--auto]	Resolve spec-audit findings one by one with summaries, alternatives, a recommendation, and SIW issue creation.
/trackbed:siw:transfer-to-linear	User	[siw-dir] [--project <name-or-id>] [--team <team>] [--dry-run] [--skip-done] [--skip-existing|--retry]	Move a planned SIW initiative into Linear for good: one project with its documents, milestones, and issues. One-way.

Discovery, Research & Learning

Requirements discovery, primary-source research, and checks that you understand a topic.

Skill	Invocation	Arguments	Description
/trackbed:research	User, Auto	[research question or topic] [--output <path>]	Answer a question from primary sources, such as official docs, specs, or source code, and save one cited Markdown artifact.
/trackbed:discovery:wayfinder	User	[initiative description | map-path [ticket-id]]	Chart a huge or foggy initiative as a local decision map and resolve one open question per session until it is ready for planning.
/trackbed:discovery:interview	User, Auto	[topic | file-path(s) | siw] [--siw] [--apply] [--ideate] [--decision-tree] [--research]	Interview you about a topic, proposal, or spec to uncover requirements, priorities, and non-goals, then write a plan; --siw feeds an SIW spec.
/trackbed:discovery:strategic-inquiry	User	[focus, e.g. 'onboarding', src/auth, or omit for whole repo] [--max-questions N] [--output <path>] [--inline]	Generate ranked strategic questions about hidden assumptions, contradictions, and load-bearing decisions in a codebase or area.
/trackbed:learn:verify-understanding	User	[topic: PR, branch, feature, document, spec, etc.] [--answer-options|--choices]	Confirm that you understand a PR, feature, spec, or other topic through questions and a checklist before calling it done.

Product

Product direction, demand validation, post-prototype PRDs, behavior documentation, and design critique.

Skill	Invocation	Arguments	Description
/trackbed:product:strategy	User	[optional: section or notes to revisit, e.g. 'metrics', 'active tracks']	Create or update STRATEGY.md, a concise product anchor covering problem, users, metrics, active tracks, and non-goals.
/trackbed:product:validate-demand	User, Auto	[idea or evidence] [--output <repo-relative-path>]	Judge whether a product idea has enough demand evidence to justify more work: GO, PIVOT, KILL, or INSUFFICIENT EVIDENCE.
/trackbed:product:prd	User	[feature name or path to an existing PRD]	Write a PRD after prototyping that records which candidate ships and why, how success is measured, and when to roll back; log results after launch.
/trackbed:product:describe-behavior	User	[product or surface] [--source <path>] [--output <path>] [--resume]	Document how an existing product behaves from the user's perspective, grounded in code, tests, and the running app.
/trackbed:product:design-critic	User, Auto	[file-path, screenshot, URL, or product question]	(experimental) Critique product UI and UX, including flows, hierarchy, trust, and jobs-to-be-done, rather than visual polish alone.
/trackbed:product:review	User	<url|auto> [--flows <flow1,flow2,...>] [--focus <dimension>] [--inline]	(experimental) Review a live product across flows for navigation, discoverability, onboarding, consistency, dead ends, and trust.

Linear

Linear issue tracking. These skills need an authenticated Linear MCP server; issue-to-pr and review-pr also use the GitHub CLI. See the shipping guide.

Skill	Invocation	Arguments	Description
/trackbed:linear:backlog-refine	User	[team] [--project <name>] [--label <name>] [--stale-days <n>] [--limit <n>] [--apply]	Grade a Linear team's open backlog and propose completing, cancelling, merging, splitting, or rewriting issues. Read-only unless --apply.
/trackbed:linear:breakdown-findings	User	[--auto] [--ask] [--dry-run] [--resume <partial-report-path>] [--team <team>] [--project <project>] [--label <label>] [--] [source ...]	Turn reviewed audit, review, scan, or QA findings into coordinated, PR-sized Linear issues.
/trackbed:linear:issue-define	User, Auto	[--auto [--ask]] [--] [issue-id or description and/or file paths for context]	Create or improve a Linear issue that fits one PR; larger work becomes a parent with sub-issues. --auto drafts from rough input.
/trackbed:linear:issue-implement	User, Auto	<ISSUE-ID> [--auto] [--set-in-progress]	Implement one Linear issue: branch setup, reference mapping, planning, coding, and fast verification.
/trackbed:linear:issue-to-pr	User	<ISSUE-ID> [--continue] [--strict] [--cycles <1-5>] [--depth <balanced|full>] [--adversarial-review] [--launch-confirmed] [--ship]	Take one Linear issue through implementation, review convergence, and verification; --ship also opens the PR and fixes CI.
/trackbed:linear:next	User	[team or selection query] [--state <name>] [--agent-ready] [--interest <preference>] [--mine|--unassigned|--both] [--project <name>] [--label <name>] [--count <n>] [--limit <n>]	Find up to 10 valuable, unstarted Linear issues by default, with plain-language summaries, when deciding what to pick up next.
/trackbed:linear:review-pr	User, Auto	[PR-number|PR-url] [ISSUE-ID]	Check that an existing PR implements its Linear issue, tracing each requirement to diff, code, and test evidence. Read-only.

Pull Requests

PR creation, review, iteration, and resolution. See the review guide and the shipping guide.

Skill	Invocation	Arguments	Description
/trackbed:pr:create	User, Auto	[--manual] [--draft] [--rebase] [--base <branch>] [--linear-issue <ISSUE-ID>] [--independent-review-line <line>] [--require-generated-description] [--authorize-history-rewrite]	Publish a new PR unattended: narrative commits, a generated title and body, and a lease-protected first push; --manual keeps the prompts.
/trackbed:pr:stack	User	[init [--base <trunk>] <branches...> | submit | sync | merge <pr-number> | adopt [--base <trunk>] <branches...> | status]	Create and manage GitHub stacked PRs with gh-stack when a change should land as a chain of small dependent PRs.
/trackbed:pr:adversarial-review	User, Auto	[--provider claude|codex] [--model <id>] [--base <branch>] [--requirements <authoritative requirements>]	Get one read-only review of your branch from a different model provider than the host you are using.
/trackbed:pr:code-review	User, Auto	[--subagent-model <model>] [aspects] [--emphasize <dim>...] [--base <branch>] [--previous-review <path>] [--no-cleanup] [--parallel] [parallel] [--team] [--inline] [--loop] [--no-diff-comments]	Review branch changes for bugs, tests, errors, types, architecture, security, performance, and cleanup; writes REVIEW_OVERVIEW.md.
/trackbed:pr:review-convergence	User, Auto	[--subagent-model <model>] [--strict] [--rounds <1-5>] [--depth <balanced|full>] [--adversarial-review [--adversarial-provider claude|codex] [--adversarial-model <id>]] [--derive | LINEAR-ISSUE | --requirements <authoritative requirements>]	Take a clean committed branch through ordered review gates, bounded fixes, and verification against frozen requirements.
/trackbed:pr:resolve-review	User, Auto	[--team] [--implement-only] [--granular] [--severity ...] [--source local|online|conductor] [review|url|instructions]	Implement fixes for code-review or Conductor diff-comment findings as commits on the current branch.
/trackbed:pr:github-review-reply	User, Auto	[--auto] [--implement|--no-implement] [--post] [--resolve] [--inline] [--humanize] [--human-only|--bot-only|--include-bots] [--all] [--only <login>] [pr-url|instructions]	Triage review comments on your own PR, make the needed changes, and draft or post replies.
/trackbed:pr:github-review	User	[--subagent-model <model>] [pr-number|pr-url] [--stack] [--draft-review] [--no-clarity-check] [--base <ref>] [--categories a11y,ux,product,visual] [--code-only] [--sequential] [--fresh] [--full] [--include-bots] [--all-threads] [--inline] [--keep-worktree]	Review someone else's GitHub PR or its whole stack, including linked Linear issues. Report-only unless you ask for a pending review.
/trackbed:pr:review-queue	User, Auto	[--repo OWNER/NAME] [--since ISO-8601] [OWNER/NAME#N ...]	See which PRs wait on your review and where each stands: your part, CI, merge state, and linked issue; rechecks report what changed. Read-only.
/trackbed:pr:fix-ci	User, Auto	[--fixup] [--manual] [--no-consolidate] [--rebase|--no-rebase] [--scope-plan <archived-plan>]	Repair failing CI checks and reviewer feedback on an open PR with targeted pushes until checks pass; --manual keeps the prompts.
/trackbed:pr:shape-up	User	—	Get an open PR ready to merge: rebase when behind its target, fix CI until checks are green, and fix and answer AI reviewer comments. Does only what is needed.
/trackbed:pr:generate-description	User, Auto	[--manual] [--no-update] [--check [--strict]] [--base <ref>] [--base-commit <oid>] [--linear-issue <ISSUE-ID>] [--independent-review-line <line>]	Write a PR title and body from the diff, commits, and Linear context, updating an open PR unless --manual; --check reports drift.
/trackbed:pr:walkthrough	User	[--report [branch|commit|PR#|range] | --report --base <ref> | --base <ref>] [--output <path>]	Generate a visual HTML walkthrough of a branch or PR; --report produces a shareable before/after architecture report.
/trackbed:pr:gut-check	User, Auto	[--base <branch>] [--intent <text>]	Get a fast first-reader reaction: does anything on this branch look strange, unusual, or unnecessary?
/trackbed:pr:triage-risk	User, Auto	[--base <branch>]	Decide how much review the branch needs: a low, medium, or high risk tier and the review commands to run. Read-only.
/trackbed:pr:convention-review	User, Auto	[--subagent-model <model>] [--base <branch>] [--threshold 0-100] [--baseline <path>] [--inline] [--no-diff-comments]	Flag drift from documented rules and established codebase practice, plus overcautious code, with evidence for every finding.
/trackbed:pr:overengineering-review	User, Auto	[--subagent-model <model>] [--base <branch>] [--inline] [--no-diff-comments] [--requirements <text>]	Flag speculative generality and unlikely edge-case handling that the requirements do not need.
/trackbed:pr:product-review	User, Auto	[--subagent-model <model>] [app-url|auto] [--categories a11y,ux,product,visual] [--threshold 0-100] [--base <branch>] [--parallel] [--team] [--inline] [--no-diff-comments]	Review the product value, UX, visual consistency, and accessibility of your changes with specialist agents.
/trackbed:pr:rebase	User, Auto	[--manual] [--force-push] [--fix-ci] [--base <branch>] [--all-open]	Rebase onto the PR's target branch and force-push with a lease, unattended by default; --force-push or --manual keeps the gates.

Code Changes

Plan-driven implementation, migrations, deprecations, design and security guidance, and cleanup.

Skill	Invocation	Arguments	Description
/trackbed:code:plan-to-pr	User	<attached plan | PR_PLAN_W##L_*.md> [--continue] [--strict] [--cycles <1-5>] [--depth <balanced|full>] [--ship]	Implement one self-contained PR_PLAN_*.md on a fresh branch, then review, verify, and optionally ship it as a PR.
/trackbed:code:work-from-plan	User, Auto	[plan path | inline plan]	Execute one standalone implementation plan, such as a PR_PLAN_*.md file or a pasted checklist, as bounded work on the current branch.
/trackbed:code:breakdown-findings	User	[--auto] [--resume|--reconcile|--split] [--all | plan-file ...] [--worktree <path>] [--source <ref>] [--base <ref>] [--] [source ...]	Cluster validated review, audit, or QA findings into PR-sized PR_PLAN_*.md files; --split plans slices of a branch and --reconcile updates plan sets.
/trackbed:code:api-design	User, Auto	[--design-twice]	(experimental) Design stable APIs and module boundaries for endpoints, public modules, and SDKs; --design-twice compares alternative shapes first.
/trackbed:code:harden-security	User, Auto	—	Apply secure defaults when code handles untrusted input, authentication, sessions, dependencies, or personal data.
/trackbed:code:migrate	User	<target e.g. 'Angular 19', 'React 19', 'Node 22'> [--auto]	(experimental) Plan and run framework or library version upgrades, such as Angular, React, or Node, in phases with verification gates.
/trackbed:code:deprecate	User	—	Plan and run the deprecation of code, features, APIs, or persistent data shapes: build the replacement, announce, migrate incrementally, then remove.
/trackbed:code:optimize	User	[spec.yaml | optimization goal] [--auto]	(experimental) Run repeatable, metric-driven optimization experiments when a measurable outcome needs variants instead of guess-and-check.
/trackbed:code:refactor-pass	User	[scope ... | --rewrite]	Simplify recent changes and clean up AI slop slice by slice without changing behavior; --rewrite redoes mediocre recent work properly.
/trackbed:code:encode-constraint	User	[mistake description | path:line example | review comment]	Turn an observed agent mistake or review nit into a type, lint rule, or check that rejects it, proven to fail on the mistake and pass on the repository.

Audits

Codebase, dependency, test-suite, and agent-practice audits that report findings without changing code.

Skill	Invocation	Arguments	Description
/trackbed:code:refactor-opportunities	User, Auto	[--subagent-model <model>] [full | pr | path <file-or-folder> | feature <name>] [--for <spec-file | description>] [--ranked [--solo] [--output <path>] [--max-findings N] [--base <ref>]]	Find refactoring candidates, tech debt, and implicit state machines in a repo, PR, path, or feature, or before a change (--for); --ranked ranks weaknesses.
/trackbed:code:outside-view	User	[--subagent-model <model>] [--raters N] [--output <path>] [--compare <report-path>] [--cross-model]	Rate a codebase with unprimed raters to surface unknown-unknown quality problems, especially in AI-engineered code.
/trackbed:code:agent-readiness	User	[--subagent-model <model>] [--auto]	Score how well a codebase suits AI coding agents across five dimensions and get a prioritized refactoring plan.
/trackbed:code:audit-agent-config	User, Auto	[scope-path]	Audit the repository's agent instructions, skills, hooks, MCP servers, and permissions for redundant, outdated, or conflicting entries. Read-only.
/trackbed:code:audit-practice	User	[--recent N] [--no-sessions]	Check how you work with coding agents against 18 common mistakes, from effort and alignment to review and steering, and get the fixes to make first.
/trackbed:code:audit-security	User	[--output <repo-relative-path>]	Audit a repository's security posture and attack surfaces before remediation, producing an evidence-ranked report with coverage gaps.
/trackbed:code:copy-review	User, Auto	[--pr] [--base <branch>] [--threshold 0-100] [--inline] [--all | <scope-path>]	Find unnecessary, redundant, or duplicated UI text across the codebase, or in the branch diff when run on a feature branch.
/trackbed:deps:audit	User	[--auto]	(experimental) Audit dependencies for outdated packages, vulnerabilities, and staleness, and get a prioritized upgrade plan.
/trackbed:test:audit	User, Auto	[full | path <file-or-folder> | changed [--base <ref>]] [--max-findings N]	Find low-value, brittle, obsolete, or duplicated tests and get a read-only report of what to remove, repair, or consolidate.

Git

Git history management and commit operations. History rewrites are covered in the shipping guide.

Skill	Invocation	Arguments	Description
/trackbed:git:fixup	User	[--skip-tests|--skip-build|--skip-lint|--skip-all] [--no-confirm] [--base=<branch>] [instructions]	Fold unstaged changes into the commits they belong to on the current branch, validate, and autosquash.
/trackbed:git:recreate-commits	User, Auto	[--manual] [--coarse|--granular] [--base <branch>] [--base-commit <oid>] [--backup-ref <branch>] [--after <commit>] [--force-backup] [--require-unstacked] [--no-push] [--authorize-history-rewrite]	Rewrite unpublished branch history into narrative commits behind a backup ref, unattended unless --manual. Not for merged or shared branches.

Debug

Bug investigation and root cause analysis.

Skill	Invocation	Arguments	Description
/trackbed:debug:find-sibling-bugs	User, Auto	[--base <branch>] [--intent <text>]	Use a bug-fix branch as a worked example and search the codebase for other places with the same flaw.
/trackbed:debug:investigate	User	[bug description, error message, or issue reference] [--auto]	Investigate a bug step by step: reproduce it, isolate it, trace the root cause, and fix it.
/trackbed:debug:triage-to-issue	User	[bug description, error message, or Linear/SIW issue ref] [--yes | --auto]	(experimental) Turn a bug into an implementation-ready Linear or SIW issue with root-cause evidence and a red-green TDD fix plan.

Browser & QA

Live product inspection and structured testing.

Skill	Invocation	Arguments	Description
/trackbed:browse	User, Auto	<url|auto> [--screenshot] [--console] [--network]	(experimental) Drive a live app through whichever browser MCP is available: navigate, screenshot, interact, and capture evidence.
/trackbed:qa	User, Auto	<url|auto> [quick|diff-aware|targeted <route>] [--base <branch>] [--regression] [--inline] [--legacy-console]	Test a running app with smoke, diff-aware, or targeted checks and get QA_REPORT.md with screenshots, repro steps, and severity.
/trackbed:qa:intake	User	[optional starting context]	Log several bugs from a manual QA pass as Linear or SIW tickets through a light, quick conversation.

Visual

Generate and inspect visual output, including self-contained HTML pages, UI checks, and diagrams.

Skill	Invocation	Arguments	Description
/trackbed:visual:check-slop	User	<file-or-directory> [--fix]	Run a deterministic check for AI-style UI patterns in HTML screens; --fix applies safe fixes.
/trackbed:visual:explain	User, Auto	[topic or description | --onboarding [focus-area or audience]]	Generate a self-contained HTML page that explains a system, change, plan, or data visually; --onboarding builds a codebase guide.

Documentation & Writing

Decision records, specs, solution notes, document review and conversion, and plain-language editing.

Skill	Invocation	Arguments	Description
/trackbed:docs:adr	User, Auto	[decision title]	Write an Architecture Decision Record for a significant, long-lived decision, following the project's existing ADR format.
/trackbed:docs:feature-spec	User	[feature name, brief description, or origin plan path] [--synthesize|--auto]	Write a lightweight PRD-style feature spec before coding, for features that need written alignment but not the full SIW workflow.
/trackbed:docs:track-rejected-enhancements	User	<record|check|append|reconsider> <concept>	(experimental) Record rejected enhancement ideas in .out-of-scope/ and check new requests against them.
/trackbed:docs:review	User	[markdown-path] [--trade-offs] [--inline|--file|--output <path>]	Review one Markdown document, such as a plan, proposal, or README draft, and get a verdict with severity-ordered findings.
/trackbed:docs:solution-note	User	[problem, lesson, or context] | --refresh [solution-note-path|--all] [--apply]	Record a solved problem in docs/solutions/ after a fix, migration, or lesson; --refresh audits existing notes against the code.
/trackbed:docs:sync-release	User	[<base>...<ref>|--base <rev> [--ref <rev>]|--current] [--apply]	Compare install, usage, architecture, and contract docs against a release range and report drift; --apply fixes it.
/trackbed:docs:to-markdown	User, Auto	—	Convert PDF, Office, HTML, image, audio, and other files to Markdown with MarkItDown.
/trackbed:docs:ubiquitous-language	User	—	Extract a DDD-style glossary of domain terms from the conversation into UBIQUITOUS_LANGUAGE.md, flagging ambiguous terms.
/trackbed:docs:update-agents-md	User	—	Add or update guidelines in AGENTS.md, such as a no-compatibility-code policy, in a structured, keyword-based style.
/trackbed:text:clarify	User, Auto	[--humanize] [file-path or text]	Rewrite prose so readers can understand and act on it quickly; --humanize removes AI-writing patterns instead.

Skill Authoring

Create and review plugin skills.

Skill	Invocation	Arguments	Description
/trackbed:skill:create	User	[skill-name or description]	Scaffold a new plugin skill with best-practice structure, frontmatter, and progressive disclosure.
/trackbed:skill:review	User, Auto	[skill-path | skill-name | proposed skill text]	Review a skill for focused scope, progressive disclosure, portability, safety, and documentation quality.

Workflow & Configuration

Environment checks, verification, next-skill suggestions, artifact cleanup, guided setup scripts, session review, and hook configuration.

Skill	Invocation	Arguments	Description
/trackbed:setup	User, Auto	[--json|--help]	Run a read-only health check of the tools, repository context, plugin version, and workspace this plugin depends on.
/trackbed:verify:run	User, Auto	[--fast | --full] [--force] [--base <branch> [--base-commit <oid>]]	Run the project's tests, linting, formatting, builds, and type checks for affected code; --fast runs the quick tier.
/trackbed:workflow:cleanup-artifacts	User, Auto	[--auto | --siw]	Delete disposable workflow artifacts such as review, audit, and QA reports, plans, and diagrams; --siw retires the siw/ workspace.
/trackbed:workflow:suggest-next	User, Auto	[question]	Suggest at most three trackbed commands for a task or for the artifacts in your workspace, read from the installed skills. Read-only.
/trackbed:workflow:wizard	User, Auto	[procedure description or target script path]	Generate an interactive Bash wizard for manual procedures such as third-party setup, one-off migrations, or CI secrets.
/trackbed:hooks:toggle	User	<status|reset|hook-name> [enable|disable]	Enable, disable, list, or reset this plugin's hooks.
/trackbed:session:automate-repeats	User	[session-paths or --recent N] [--effectiveness] [--create|--auto]	Review recent agent sessions for repeated work and friction, and suggest skill improvements or new automation.

Background Skills

Auto-triggered by the host based on context. These don't appear in the / menu.

Skill	Invocation	Arguments	Description
trackbed:git:commit-message	Background	—	Write commit messages that follow the repository's commit convention, with names-only secret and stray-file checks and AI-attribution rules.