gidde032

agentic-collaboration-cadence

Establish and operate the human-in-the-loop authority model for agentic project work: distinguish product ratification, implementation, review repair, and merge or release authority; define an authorization envelope; continue pre-authorized repository and GitHub work with exception-only pauses; consolidate useful checkpoints; surface judgment calls, cost tripwires, and honest evidence. Use when planning substantial work, calibrating agent autonomy, deciding whether an Issue, PR, merge, deployment, publication, or repository-visibility action needs approval, structuring a review-and-repair cycle, or correcting a workflow that asks for approval too often or assumes too much authority. Triggers: "human in the loop", "authorization boundary", "implementation approval", "ratified but not approved", "exception-only pause", "work in the background", "checkpoint", "autonomous execution", "impact-ranked plan", "pick the slice", "judgment call", "how often should the agent check in".

agentic-docs-github-migration

Explicitly invoked migration workflow for an existing solo-maintained repository whose documentation, roadmap candidates, shipped history, and GitHub governance need consolidation. Use only when the user invokes `$agentic-docs-github-migration profile`, invokes `$agentic-docs-github-migration execute` with an absolute profile path, or explicitly requests this named repository-wide migration. Do not trigger for ordinary documentation edits, issue creation, CI changes, feature work, project planning, new-project bootstrap, or mid-project maintenance.

agentic-driving-weaker-models

Operating manual for getting strong-model-quality results out of cheaper/smaller agent sessions, based on observed failure modes from validation dry runs and two projects' review history. Use when: choosing which model tier runs a task; composing a brief for a cheaper session; a session keeps making a characteristic mistake (trusting docs without verifying, fixing the symptom not the contract, skipping gates under constraint); deciding whether to escalate to a stronger model or the human; a cheaper agent wants to self-check its own known weak spots before delivering. Covers the gap taxonomy, the mitigation matrix, the task→tier routing table, brief-composition checklists, and escalation triggers. NOT for reviewer-brief mechanics (agentic-review-orchestration) or raw cost math (agentic-session-economics).

agentic-implementation-orchestration

Orchestrate an approved Issue-backed implementation with a strongest-capability parent and bounded implementer agents. Use when decomposing an authorized feature or phase; mapping work packets to Issue acceptance criteria; deciding what the parent retains versus delegates; assigning exclusive files or modules; preserving unrelated work in a shared workspace; integrating changes into coherent commits; opening or updating a draft PR after local gates; or independently accepting delegated work before cold review. Covers dependency graphs, capability routing, ownership, concurrency, compact and full work packets, structured returns, escalation, integration, tracking impact, and parent-side verification. Do not use to decide product scope, treat Issue ratification as implementation authority, run cold review, or perform ordinary Issue/PR administration.

agentic-phase-workflow

Run the issue-backed development loop that turns a ratified spec into shipped increments: select an owning Issue, confirm implementation authority, implement on a focused branch, open a draft PR after the first coherent green commit, review independently, repair with regression tests, reconcile documentation and tracking, and close delivery state after merge. Use when planning or running a feature phase, hardening pass, bundled phase, spec-drift audit, workflow-improvement release, or release closeout; when deciding whether work is phase-shaped or pass-shaped; or when sequencing Issues within a milestone. Do not use for day-one project bootstrap, repository-wide documentation migration, or ordinary Issue/PR administration that does not require choosing or running a development shape. Premise: the workflow is the product, as much as the code.

agentic-product-discovery

Turn an uncertain product or feature idea into a ratified product-and-design contract before architecture and implementation. Use when deciding what to build; comparing existing software with a custom product; clarifying users, workflows, scope, value, requirements, and the capture of deferred ideas (their placement belongs to `agentic-project-memory`); shaping information architecture or interaction models; exploring UI directions, palettes, typography, density, navigation, components, or branding; iterating through visual mockups and synthesis rounds; or preserving approved product and design decisions for fresh agents. Routes rendering, image generation, system design, project bootstrap, and implementation to their owning skills.

agentic-project-bootstrap

The day-one instantiation runbook for the whole human-directs-AI-agents workflow system. Use when starting a new project; when adopting this workflow on an existing project; for day-one setup; when standing up a spec, memory files, GitHub-backed planning and delivery, quality gates, a review cadence, and an evolution ledger together; when someone says "set up the workflow", "bootstrap this repo", "start a project the agentic way", or "instantiate the agentic workflow". This is the ROUTER — it sequences the sibling skills and gets out of the way. Do not trigger it for ordinary mid-project Issue, PR, documentation, or implementation work. Premise: the workflow is the product, as much as the code.

agentic-project-memory

Maintain durable project memory and assign each kind of information to one authoritative location across sessions, agents, GitHub Issues, milestones, pull requests, and source-of-truth repository documentation. Use when creating or updating a handoff; deciding whether information belongs in an Issue, PR, spec, roadmap, changelog, task file, practices catalogue, retrospective, FAQ, or pending lessons; deciding where a deferred idea is recorded; reconciling documentation drift; reconciling which documents a completed phase must update; preserving session continuity; documenting tooling failures; splitting an oversized internal document; or defining documentation-impact accounting. Do not use for ordinary GitHub Issue or PR operations that do not affect project memory or source-of-truth boundaries, or for writing ordinary reference documentation -- a README, API docs, docstrings -- that carries no source-of-truth status.

agentic-review-orchestration

Run independent, contextless multi-agent review against an exact draft-PR diff and accepted Issue contract. Use after an implementation is integrated and locally green; before marking a PR ready; when running a release-hardening audit; when convergence across independent reviewers would improve confidence; when building isolated reviewer briefs; or when validating and triaging findings into approved repairs, linked deferral Issues, false alarms, and other terminal dispositions. Covers reviewer isolation, exact base/head targets, personas, orthogonal audit lenses, disposable-state verification, bounded evidence-oriented briefs, parallel versus serial review, provisional ledgers, human repair approval, and public-safe PR summaries. Do not use for implementation, ordinary PR administration, or reviewing work that does not yet have a coherent integrated target.

agentic-session-economics

Keep agent-session spend proportional to work delivered. Use this skill when: planning a new agentic workflow; noticing session costs feel disproportionate to output; deciding whether to full-read a large file or use a tail-read; weighing the cost consequences of a model tier; deciding when to start a fresh session; batching doc updates; setting up cost discipline before a project grows; fighting broken tooling in the agent layer. Triggers: "my usage is going fast", "why is this session so expensive", "which model should this subagent run on", "how do I keep costs down", "when should I start a fresh session", "how do I read large files efficiently", "token discipline". Defer tier *selection* for a given task to `agentic-driving-weaker-models`; this skill owns what a tier choice costs, not which tier fits the work. Not for API or product pricing questions.

agentic-verification-discipline

Build trustworthy evidence for agent-generated changes and make required verification enforceable locally and in GitHub. Use when converting accepted findings into fail-before-fix regression tests; defining or weakening numeric gates and coverage floors; recording red/green evidence in a PR; creating or selecting a repository-owned GitHub Actions required check and its check source; verifying ruleset enforcement; preventing unexpected skips, enforcing a skip budget, or pinning a smoke inventory; testing stable documentation promises; synchronizing tool ignore lists; deciding where a regression test belongs; or diagnosing a local-versus-CI difference where CI enforces a gate the local run does not. Do not use for ordinary CI log debugging that changes no verification contract, for designing a test plan or test strategy that carries no gate or enforcement change, or for deciding product scope and finding disposition.

agentic-workflow-evolution

Improve a human-directs-AI-agents workflow through evidence rather than theory. Capture and route workflow observations; distinguish project-local lessons, cross-project candidates, and plugin-wide practices; design real-work experiments; evaluate reviewer, CI, migration, memory, gate, or session failures; promote, reshape, or retire practices; and reconcile the Issues, PRs, rules, practices, skills, and plugin releases affected by an approved change. Use when deciding whether a lesson belongs in pending-lessons, practices, a GitHub Issue, project rules, or a reusable plugin; when updating a skill; when a recurring failure suggests a workflow change; or when auditing whether current guidance is still supported. Triggers: "should we change the workflow", "promote this lesson", "update the plugin", "workflow experiment", "named failure mode", "why did this recur", "is this practice reusable", "retire the old practice", "log this revision", "what evidence supports this rule".