The skills
Describe the work. Your agent chooses the methods the task needs. Installation makes a skill available; it does not require the agent to use every installed method.
Skills are methods and disciplines#
The default roster includes planning, implementation, review, verification, communication, and engineering disciplines. Explicit exclusions are supported; diagram-design and architecture-map are optional asset integrations.
| Work | Some relevant methods |
|---|---|
| Establish continuity | project-router, groundwork |
| Resolve uncertainty | grilling, grill-with-docs, wayfinder, domain-modeling |
| Gather evidence | research, prototype, control-cli, control-ui |
| Describe and implement work | to-spec, to-tickets, implement, tdd |
| Review and verify | delivery-review, review-lens, ponytail-review, verify-this |
| Diagnose failures | diagnosing-bugs |
| Exercise engineering restraint | ponytail, codebase-design, boundary-discipline, type-system-discipline, model-the-domain |
| Reason about concurrency | concurrency-method |
| Examine architecture or tests | improve-codebase-architecture, sweep-tests |
| Explain and communicate | answer-plainly, technical-writing, unslop, show-me, progress-check |
| Retain product knowledge | product-description, record-architecture-decisions, what-to-build |
| Protect evidence and secrets | fail-loud, airgap-secrets |
The review runs as a subagent the harness starts fresh. When a session will not start one (some Claude Code sessions decline delegation unless asked), the agent leaves the ticket implemented and blocked for review with the reason recorded, and the owner opens a fresh session to review it; nothing is skipped silently.
These are neighboring methods, not a fixed order or a promise that every method is installed. The installed roster is the availability list; exclusions still win. Every skill has one class, listed with the roster in AGENTS.md: an entry (project-router), a stage of the delivery arc, a support skill a stage dispatches, a discipline the agent applies on its own judgment, or a situational skill that fires on a situation in the conversation and is offered in one line when its output is a durable file you did not ask for. Codex's per-method allow_implicit_invocation: false disables implicit discovery for stages, not work already authorized by the owner; Claude Code reads the same conduct rules in the block. Neither creates a separate stage approval.
The installed descriptions and project-router's ROUTING.md, the pipeline map, help distinguish them. You do not have to select a persona or call each stage yourself. Each stage or support skill ends with a Handoff section naming what it consumes, produces and hands to next; each situational skill with a line naming its durable output.
Engineering knowledge is garden's#
The skills are methods: how to work. Engineering knowledge, what good looks like for a given language, stack or concern, is garden's, a separate product. greenline reaches it only through the optional garden connector. With the connector enabled, one more skill, use-garden, is installed; it tells the agent when a consultation is relevant and how to make one. Without the connector, no knowledge is read. The garden connector explains the consultation and its receipt.
Your decisions stay settled until you or the authorized agent deliberately revise them. A root statement in .greenline/DECISIONS.md can exclude knowledge for one root; the exact JSON format and its two exclusion forms are in the installed WORK.md. An exclusion governs that exact root, even when something points at an excluded unit, and it does not affect sibling roots.
Where the methods come from#
Preserved methods come from Matt Pocock, Dietrich Gebert's Ponytail, Cursor's team kit, pstack, Freddie Northam, Steve Ruiz, Eric Clemmons, HumanLayer, Cathryn Lavery, Alaa Mendili, and tjcages. Source pins and notices remain recorded. greenline does not imply endorsement.
Preserved methods are edited copies of their sources: the steps and support are the author's, and every difference (homes, names, handoffs, harness mechanics, scope) is recorded with its reason. A change to an author's own steps is made only on a ruling of greenline's maintainer and is recorded with it. The third-party notices identify each source and its terms, and the provenance pages show every difference from its source with the reason for each, rendered from the ledger.
What the ledger establishes#
Read means obtaining content; select means choosing it for this task; apply means using it with a supported outcome. Repository decisions retain lasting choices. Generated receipts and sparse account annotations keep these distinctions visible. A declaration, a captured read, and independent verification are different things.