@zalom/plastic 1.14.1 → 2.0.0-alpha.1

package/PLASTIC.md

@@ -1,334 +1,141 @@
 # Plastic: Conventions
 
-> **This file is maintained by Plastic.** It will be overwritten when the
-> plugin is updated. Do not modify it: your changes will be lost.
-> For project-specific rules, use `AGENTS.md` instead.
+> **This file is maintained by Plastic.** It is overwritten on update; do not edit it. Project
+> rules go in `AGENTS.md`. Deeper doctrine lives in the six `plastic-conventions` chapters
+> (`references/<chapter>.md`: knowledge-graph, lifecycle-and-savepoints, locks-and-worktrees,
+> completion-and-done, maintenance-and-revisions, roadmaps), read on demand.
 
-Deeper doctrine lives in the `plastic-conventions` skill's chapters
-(`plastic-conventions > references/<chapter>.md`); read the one that matches your task on demand,
-it is not injected at session start.
+## Two Modes, Plus Auto
 
-## Two Processes
-
-| Process | Scope | Type | Actor |
-|---|---|---|---|
-| **Build → Observe → Repeat** | The system | Continuous loop | Coordinator |
-| **What → Why → How → Exec** | One intent | Finite lifecycle | Agent |
+| Mode | "How to execute" comes from | Who runs it |
+|---|---|---|
+| **Direct** (default) | the prompt; at most one clarifying question | this session, inline; record after |
+| **Thinking** | a conversation: each ruling an insight, then the action files | this session, then as in direct |
+| **Auto** | a registered intent with a clear prompt | a background team under a lock and a worktree |
 
-B→O→R is the Coordinator's heartbeat. W→W→H→E is what happens inside each intent.
-The connection: an intent's `## Insights` feeds the Coordinator's Observe phase.
+There are no tiers: there is only work. The Coordinator loop (build, observe, repeat) wraps
+every intent: an intent's `## Insights` feed the observe phase.
 
 ## What is an Intent
 
-A directory in the store containing `{ID}--{slug}.md` and optional supporting files.
-It represents a desire: something a human or agent wants to accomplish, explore, or understand.
-The unit of work in Plastic is always an intent, never a ticket.
+A directory in the store holding `{ID}--{slug}.md` and its record: a desire someone wants to
+accomplish, explore, or understand. The unit of work is always an intent, never a ticket. The
+day ledger is one too: the session intent, one per calendar day, that direct work records into.
 
 ```
-store/
-  ID--three-to-five-words/
-    {ID}--{slug}.md       # required - the intent itself
-    spec.md               # optional - specification (Why deliverable)
-    plan.md               # optional - implementation plan (How deliverable)
-    checklist.md          # optional - execution registry (How deliverable)
-    outcome.md            # detailed result (Exec deliverable) - mandatory at a terminal (Completed and Abandoned alike)
-    actions/              # optional - real action files (How deliverable, at least one required)
-    resources/            # optional - research, references, screenshots, diagrams
-    savepoint.md          # optional - deterministic cycle-step ledger (auto-written)
-    revisions.md          # optional - append-only structural-maintenance audit trail
-```
-
-Lifecycle files (`spec.md`, `plan.md`, `checklist.md`, `outcome.md`) have defined
-roles. Supporting artifacts that aren't lifecycle deliverables (research reports,
-reference docs, external API snapshots, screenshots, diagrams) go in `resources/`.
-Name files inside as `{type}--{description}.md` (e.g., `deep-research--gsd-core.md`).
-
-## Frontmatter
-
-Identity and knowledge graph only. Nothing operational.
-
-```yaml
----
-id: "4a1"
-intent: "Short description of the desire"
-sources: ["4a"]        # direct ascendants: intents this was created from
-chain: ["4a1a"]        # forward: what this spawned and related successors
-created: 2026-05-29
-author: human          # human | agent-name
-tags: [plastic, architecture]
----
+store/ID--three-to-five-words/
+  {ID}--{slug}.md   # required: Intent, Context, Outcome, Insights, Links sections
+  checklist.md      # while working: one item per executed request
+  savepoint.md      # while working: one line per milestone, append-only
+  spec.md           # backfilled at intent end: requests and decisions
+  plan.md           # backfilled: what was thought
+  actions/          # backfilled: how it was done (at least one ACTION_N.md)
+  outcome.md        # backfilled: what changed; mandatory at Completed and Abandoned
+  resources/        # {type}--{description}.md
+  revisions.md      # optional: append-only maintenance audit trail
 ```
 
-- Context contract: load `sources` strongly (they are what the intent was built from);
-  traverse `chain` lightly for discovery. See
-  [`how-plastic-sources-and-chains-intents.md`](https://github.com/zalom/plastic/blob/main/docs/concepts/how-plastic-sources-and-chains-intents.md) for the full model.
-- IDs use Luhmann's alternating convention: `1` → `1a` → `1a1` → `1a1a`
-- Multiple branches increment: `1a`, `1b`, `1c`
-
-See `plastic-conventions > references/knowledge-graph.md` for the linking doctrine: the tiers
-of influence (sources/chain/tags) and the `## Links` projection rules.
+Frontmatter is identity and knowledge graph only, nothing operational: `id`, `intent`,
+`sources` (direct ascendants, loaded strongly), `chain` (what it spawned, traversed lightly),
+`created`, `author` (`human` or an agent name), `tags`. Only the day ledger adds `mode: direct`.
+IDs follow Luhmann's alternating convention (`1`, `1a`, `1a1`, `1a1a`); siblings increment
+(`1a`, `1b`).
 
-## Lifecycle Stages
+## The Record: Stages as its Shape
 
-| Stage | Section | Deliverable | Detail |
-|-------|---------|-------------|--------|
+| Stage | Section | Deliverable | Skill |
+|---|---|---|---|
 | **What** | `## Intent` | `{ID}--{slug}.md` | `plastic-intent-creating` |
-| **Why** | `## Context` + Decisions | `spec.md` | `plastic-intent-brainstorming` enriches Context/Decisions; `plastic-intent-speccing` writes `spec.md` |
-| **How** | Planning | `plan.md` + `actions/ACTION_N.md` (at least one) + `checklist.md` | `plastic-intent-planning` |
-| **Exec** | Execution | `outcome.md` | `plastic-intent-executing` |
+| **Why** | `## Context` + Decisions | `spec.md` | `plastic-intent-speccing` (thinking mode) |
+| **How** | `checklist.md` + `actions/` | `plan.md` | `plastic-intent-speccing` writes the action files |
+| **Exec** | `## Outcome` | `outcome.md` | `plastic-intent-executing`; `plastic-intent-ending` closes |
+
+Stages are the shape of the record, not checkpoints: nothing blocks a write. Only "how to
+execute" (a checklist item plus its action) must exist before work; the rest is backfilled at
+intent end from what was recorded while working.
 
 Invoke a skill for your harness: Claude Code uses the slash form (`/plastic-intent-creating`);
 Codex CLI uses a dollar prefix instead (`$plastic-intent-creating`), and may also select a skill
-implicitly by matching its description. Skill names elsewhere in this document are given bare
-(`plastic-intent-creating`); add the prefix for your harness.
-
-`## Insights` is the append-only log of durable discoveries captured throughout ALL stages.
-An insight is a discovery worth keeping for later reads: novel, or old but newly relevant,
-surfaced at any stage (What, Why, How, Exec). It is the most interesting residue of an
-intent, the part a future reader most wants. **Append-only means newest entry at the bottom;
-never prepend.** This ordering is a hard convention: Insights are the semantic trace of an
-intent, and a consistent newest-last order keeps that trace readable across every intent.
-
-Every entry leads with a fixed, machine-parseable prefix `{utc-iso8601} · {stage} · {author}`,
-for example `2026-06-24T08:13:05Z · Why · plastic-brainstorming (autonomous)`. The UTC ISO8601
-timestamp (to the second, trailing `Z`) is the same convention the savepoint ledger uses, so
-the store has one timestamp convention. This per-entry prefix is not prepending the entry:
-entries stay append-only, newest at the bottom; the prefix only stamps each line with when,
-which stage, and who.
-
-The blessed write path is the `insight-append` helper
-(`scripts/insight-append <intent_dir> <text> --stage S --author A`), which ships with every
-install and update, formats the prefix, validates it, and appends at the bottom. Hand-editing
-`## Insights` is an escape hatch; the helper is the default so the format cannot drift.
+implicitly by matching its description.
 
-See `plastic-conventions > references/lifecycle-and-savepoints.md` for the subagent
-report-home contract (how an insight reaches the intent when the writer cannot write the file
-itself) and `savepoint.md`'s role.
+`## Insights` is the append-only log of durable discoveries from every stage, newest at the
+bottom, never prepended, each entry prefixed `{utc-iso8601} · {stage} · {author}`. Write
+through `scripts/insight-append <intent_dir> <text> --stage S --author A`, which ships with
+every install and update, formats the prefix, validates it, and appends at the bottom.
 
-## Tiers (proportional auto sizing)
+## Agents and Skills
 
-Auto mode sizes every intent S/M/L at Why: S = single mechanism or file cluster (hours);
-M = one subsystem (about a day); L = cross-cutting or novel design.
+An auto team is `plastic-enforcer` (the lead) plus `plastic-executor`; models live in
+`agents.models.<harness>.<agent>`.
 
-Tier is recorded as a `Tier: S|M|L` line at the top of spec.md. It is convention-only,
-read by the orchestrator, not enforced by any gate or by doctor.
-
-Settledness is recorded separately, as a `Settled: yes (<reason>)` line directly beneath the
-`Tier:` line, above the `# Spec:` heading. It records that a design has already done its
-thinking, so a later stage can trust it without deriving it again. An absent line means not
-settled; there is no `Settled: no` form. Settledness and scope are independent, so a large
-intent can be settled and never becomes a smaller tier. `scripts/lib/spec_header.rb` is the
-only parser of both lines; nothing else reads the grammar.
-
-See `plastic-conventions > references/tiers-and-dispatch.md` for topology by tier, what never
-gets cut, and guided mode.
-
-## Agent Models and Dispatch (intent 116)
-
-| Stage | Agent |
-|---|---|
-| What | `plastic-intent-discovery` |
-| Why | `plastic-brainstorming` + `plastic-spec-specialist` |
-| How | `plastic-planner` |
-| Exec | `plastic-executor` |
-| Done | `plastic-intent-curator` |
-
-See `plastic-conventions > references/tiers-and-dispatch.md` for the advisor, model
-configuration, the dispatch contract, and the spawn preamble.
-
-## Operational Skills
-
-- **`plastic-feedback`** turns a described Plastic quirk, bug, or feature idea into a redacted
-  local report file and a prefilled GitHub issue URL; only the user submits it. If the user hits
-  a Plastic quirk, bug, or missing feature, offer to invoke the plastic-feedback skill yourself
-  instead of waiting to be asked.
-- **`plastic-doctor`** checks installation health across three scopes: core, store, and full.
-  See `plastic-doctor/SKILL.md` for the contract.
-
-## Releases and Versioning
-
-Plastic ships as versioned releases; a release is a collection of intents. Run
-`plastic-releasing` for the full flow (version bump, tag, GitHub release, npm publish).
-Deprecations are declared in `deprecations.yml` and shown at SessionStart.
-
-## Gotchas
-
-- **Artifacts go in the intent directory.** Never create `docs/plans/`,
-  `docs/specs/`, `researches/`, or similar. All meta-artifacts go in
-  `~/.plastic/store/ID--slug/` or the project store equivalent.
-- **Code goes in the project. Everything else goes in the intent.**
-  Plans, specs, checklists, savepoints: all in the intent directory.
-- **The global store is never pushed.** `~/.plastic/` is git-tracked locally
-  but contains sensitive data. Never push to a remote.
-- **Agent-created repos are private by default.** Use `gh repo create --private`.
-- **State is derived from what exists.** No `## Context` = fleeting intent.
-  `## Context` exists = permanent/actionable. `## Outcome` exists = done.
-- **Status lives on actions, not intents.** An intent can have parallel workstreams.
-- **`outcome.md` = done.** Presence signals completion. Don't write it until
-  checklist is fully checked.
-- **Delegation overrides external skill defaults.** When delegating to
-  brainstorming, writing-plans, etc., Plastic's directory rules override
-  their default output paths.
+Twenty skills, each `plastic-<name>`, listed by your harness. `plastic-doctor` checks
+installation health (core, store, and full scopes). `plastic-feedback` turns a Plastic quirk,
+bug, or feature idea into a redacted local report and a prefilled GitHub issue URL; only the
+user submits it. If the user hits one, offer to invoke the plastic-feedback skill yourself
+instead of waiting to be asked. A release is a collection of intents; `plastic-releasing`
+runs the flow. A roadmap is an ordered, delivery-side collection; `plastic-roadmap` owns it.
 
 ## State System
 
 ```
-~/.plastic/                            # Global intent store (git, never pushed)
-├── INDEX.md                           # Structure note (clusters by meaning)
-├── config.yml                         # User preferences
-├── projects.yml                       # Project slug → path registry
-└── store/                             # Strategic intents
-    └── ID--slug/
-        └── {ID}--{slug}.md
-
-~/.plastic/projects/{slug}/            # Project-scoped store
-├── INDEX.md                           # Project-scoped index
-├── AGENTS.md                          # Project stack defaults
-└── store/                             # Tactical intents
-    └── ID--slug/
+~/.plastic/                    # Global store (git, never pushed)
+  INDEX.md                     # Structure note
+  config.yml, projects.yml     # Preferences; slug -> path
+  store/ID--slug/              # Strategic intents
+  store/.sessions/<YYYYMMDD>/  # Day ledger: <YYYYMMDD>.md, checklist.md, savepoint.md
+  store/.tmp/<session>/        # current (the pointer), heartbeat; git-ignored
+~/.plastic/projects/{slug}/    # Project store: INDEX.md, AGENTS.md, roadmaps/, store/
 ```
 
-Project stores derived from `projects.yml`. No files placed in project code
-directories. SessionStart hook detects project by matching CWD.
-
-**Privacy:** Plastic is personal. All intent data under `~/.plastic/`. Each
-person has their own store. Collaboration through PRs, not shared intents.
-
-## Directory Naming
+`<session>` is the first eight characters of the session id. The pointer holds a day id (the
+day ledger takes the record) or an intent id (an auto team owns the record); session start
+writes today's day id. The SessionStart hook picks the project store by matching the working
+directory against `projects.yml`; no Plastic files land in project code.
 
-Format: `ID--three-to-five-words` (all stores).
+Naming: intents are `ID--three-to-five-words`, the file matching the directory
+(`1a1--slug/1a1--slug.md`); next id: `ruby ~/.plastic/scripts/folgezettel-id <parent_id>
+<store_path>`. Day ledgers are digits only (`20260828`), never hyphenated, never flattened to
+the store root, so the id allocator and every 1.x store walker skip them.
 
-- Root intents: sequential numbers (`1`, `2`, `3`)
-- Branches alternate: `1` → `1a` → `1a1` → `1a1a`
-- Intent file matches directory: `1a1--slug/1a1--slug.md`
-- Next ID: `ruby ~/.plastic/scripts/folgezettel-id <parent_id> <store_path>`
-
-See `plastic-conventions > references/knowledge-graph.md` for the branch-vs-root semantic
-decision: when to branch versus start a root.
-
-## INDEX.md
-
-A Zettelkasten structure note, not a table of contents. Clusters by meaning.
-
-Sections: `## Active`, `## Future`, `## Clusters`, `## Abandoned`, `## Completed`.
-
-For index maintenance, use `plastic-store-indexing`.
-
-One-line entry convention. Each index entry is ONE line: `- [<id> <terse title>](<dir>) <tags>`.
-The title is the title, not a summary: aim for about 80 characters, no multi-sentence
-descriptions. This is a self-check, not a gate.
-
-## Roadmaps
-
-A roadmap is a named, ordered, delivery-side collection of intents. Create, order, close, and
-consume one with `plastic-roadmap`; see `plastic-conventions > references/roadmaps.md` for the
-full format.
+`INDEX.md` is a Zettelkasten structure note, not a table of contents. Sections: `## Active`,
+`## Future`, `## Clusters`, `## Abandoned`, `## Completed`. One line per entry,
+`- [<id> <terse title>](<dir>) <tags>`, about 80 characters: a title, not a summary.
 
 ## Rules for Skills
 
-ALL work flows through intents.
-
-1. Before starting work, check INDEX.md for active intent. If none, create one.
-2. Skill output (spec, plan, checklist) goes in the intent directory.
-3. On completion, capture observations in `## Insights`.
-4. When done, write `outcome.md` + `## Outcome` summary. Update INDEX.md.
-5. Researches are intents. No separate folder.
-6. Intents are created only via `plastic-intent-creating`. Never hand-author an intent file. The skill self-verifies the written intent with `scripts/validate-intent` before announcing or committing, so every intent is born complete. `--intent` text is escaped for double quotes and backslashes before it lands in frontmatter, so free-form text is safe to pass as-is, and a reciprocal `chain:` append preserves the target intent's existing flow- or block-style entries.
+1. Work records into an intent: the day ledger by default, a registered intent when the
+   pointer names one. Researches are intents too; no separate folder.
+2. Artifacts go in the intent directory, never in `docs/plans/` or `docs/specs/`: code goes in
+   the project, everything else in the intent. Capture observations in `## Insights`; at the
+   end write `outcome.md` and `## Outcome`, and update INDEX.md.
+3. State is derived from what exists: `outcome.md` present means done, so never write it
+   before the checklist is fully checked. Status lives on checklist items, not intents.
+4. The global store is never pushed (`~/.plastic/` holds sensitive data); agent-created repos
+   are private by default (`gh repo create --private`).
+5. Intents are created only via `plastic-intent-creating`, never hand-authored. It
+   self-verifies with `scripts/validate-intent`, so every intent is born complete; `--intent`
+   text is escaped for double quotes and backslashes before it lands in frontmatter, and a
+   reciprocal `chain:` append preserves the target intent's existing flow- or block-style
+   entries.
 
 ## House Style (self-check)
 
-The agent is the heaviest contributor to the transcript, so terseness pays every turn. These
-are pre-send self-checks the agent applies to its own output. They are not gated.
-
-- Answer or decision first. Lead with the result, then support it.
-- Bullets over paragraphs.
-- No preamble, no end-recap. Do not restate the question or summarize what you just said.
-- One question-cluster at a time when asking the human.
-- Reasoning goes in the thinking channel, not duplicated into the visible reply. This keeps
-  the human's visibility into your reasoning without paying for it twice in the transcript.
-
-Active-intent cache rule. For the intent under active development you already hold its
-delivered artifacts in your own context: prefer revisiting that in-context memory (hit the
-cache) over re-reading them from disk, which only widens context. QMD is for OTHER or indexed
-intents, not for re-reading what you just wrote. Pairs with `/clear` plus savepoint-resume
-hygiene after each intent. Advisory self-check, not hard-verifiable.
-
-## Tabular-First Reporting (intent 160)
-
-**Default.** Tabular layout is the default shape for three surfaces: What-stage discovery
-deposits, all research reports, and all agent reporting or presentation surfaces.
-
-**Calibration.** Tables are REQUIRED for any listing or discussion of intents, and for
-explaining complex data, comparisons, or multi-factor reasoning. This is NOT a blanket
-tables-everywhere rule: simple data stays prose, and tables must not be overused.
-
-**Bullets-limit.** Use bullets only when a table genuinely does not fit the content, and
-never more than 3-5 items.
-
-**Exception.** The EM-to-CTO human briefing (`skills/auto/references/human-report-contract.md`)
-keeps its deliberate prose shape (fixed State/Risk/Call, single item, nothing to tabulate) and
-is exempt from this rule.
-
-## QMD Search
-
-QMD, Enola, and Serena are recommendations, not obligations. The deterministic entry point
-is `scripts/qmd-sync` (detect, register, reindex, status, search). Intent delivery reindexes
-the store; see `plastic-conventions > references/gates-and-enforcement.md` for gate
-mechanics.
-
-## Transition Gates
-
-| Transition | Trigger | Gate |
-|---|---|---|
-| What → Why | `spec.md` written | (none) |
-| Why → How | `plan.md` + `actions/ACTION_N.md` (at least one) + `checklist.md` | `spec.md` must exist |
-| How → Exec | Checklist has items | plan.md, checklist.md, and at least one real actions/ACTION_N.md must exist |
-| Exec → Done | `outcome.md` written | All checklist items checked |
-
-Hard blocking: hooks exit code 2 on gate failure.
-
-### The gates by name
-
-One line each. On Claude the five edit-path gates (savepoint-pre, lock-gate, code-gate,
-links-gate, create-gate) run inside one dispatcher process per Write or Edit, in that fixed
-order with the first deny winning; what each gate checks is unchanged.
-
-- **savepoint-pre** appends the savepoint ledger line before a write into an intent dir; it
-  records and never denies.
-- **lock-gate** admits only the intent's lock owner or a registered delegate to write into an
-  active intent directory.
-- **code-gate** denies a code edit before How is delivered and a code edit outside the
-  provisioned worktree (stage rule or worktree rule, first match wins), with the audited
-  `# plastic-ok` escape.
-- **links-gate** enforces the derived `## Links` projection contract at write time.
-- **create-gate** validates the proposed intent file at What write-time.
-- **bash-gate** (on the Bash matcher) intercepts a write attempted through a bash or
-  interpreter one-liner, with the same audited escape.
-- **gate-check** (PostToolUse) enforces lifecycle stage order after each write.
-- **future-intent-check** (UserPromptSubmit) surfaces parked future intents whose keywords
-  match the user's message; it informs and never denies.
-
-See `plastic-conventions > references/gates-and-enforcement.md` for the escape and logging
-detail.
-
-## Delivery Isolation and the Single-Owner Lock
+Answer or decision first; bullets over paragraphs; no preamble, no end-recap; one
+question-cluster at a time; reasoning stays in the thinking channel. Tables are the default
+for discovery deposits, research reports, and agent reports, and required for any listing of
+intents or multi-factor comparison.
 
-Exactly one session or agent develops an intent's delivery at a time. Ownership is a
-session-keyed, durable `delivery.lock` (O_EXCL) in the intent directory; liveness is a lease
-(the owner's hooks refresh the file mtime, stale means older than the TTL). The `/tmp` session
-bridge is only a cache: the lock file wins on any disagreement. A per-artifact claim token
-arbitrates the file grain underneath the lock. Every code-touching intent gets its own git
-worktree (`<repo>/.claude/worktrees/{id}--{slug}`, branch `plastic/{id}--{slug}`); code edits
-happen only inside it. On a confirmed solo delivery the lock and worktree gates relax to
-advisory; any parallel or team activity restores strict enforcement.
+## QMD, Enola, and Serena
 
-"Intent done" is one law with three signals that must agree: INDEX `## Completed` /
-`## Abandoned`, `outcome.md`, and the savepoint `Done` line; INDEX is authoritative on any
-conflict.
+Recommendations, not obligations, and only when present. QMD: prefer `qmd search` /
+`qmd query` over the `plastic-*` collections to check for existing or related intents before
+treating work as new. Enola, or Serena when Enola is absent: prefer its symbol resolution (or
+`.enola/facts.jsonl`) for code navigation over grep.
 
-Structural maintenance (WORK vs MAINTENANCE) never touches delivered content: it only records
-itself in `revisions.md`, and only proceeds when the target's `delivery.lock` is not fresh.
+## Auto Teams: the Lock and the Worktree
 
-See `plastic-conventions > references/locks-and-worktrees.md`,
-`references/completion-and-done.md`, and `references/maintenance-and-revisions.md` for the full
-doctrine.
+Locks and worktrees exist only for auto teams. One team develops an intent at a time: a
+session-keyed `delivery.lock` in the intent directory, alive while its mtime lease is fresh.
+Code edits happen in the intent's worktree (`<repo>/.claude/worktrees/{id}--{slug}`, branch
+`plastic/{id}--{slug}`). "Done" is three signals that agree: INDEX `## Completed` or
+`## Abandoned`, `outcome.md`, and the savepoint `Done` line; INDEX wins on conflict.

package/README.md

@@ -136,9 +136,8 @@ Walk this once in
 | Family | Commands |
 | ------ | -------- |
 | Mode | `plastic-tutorial`, `plastic-auto` |
-| Intent | `plastic-continuing`, `plastic-intent-creating`, `plastic-intent-starting`, `plastic-intent-brainstorming`, `plastic-intent-grilling`, `plastic-intent-speccing`, `plastic-intent-planning`, `plastic-intent-executing`, `plastic-intent-ending`, `plastic-intent-continuing`, `plastic-intent-researching` |
-| Project and delivery | `plastic-project-creating`, `plastic-project-continuing`, `plastic-roadmap`, `plastic-roadmap-continuing`, `plastic-releasing` |
-| Skill | `plastic-skill-creating`, `plastic-skill-evaluating` |
+| Intent | `plastic-intent-creating`, `plastic-intent-continuing`, `plastic-intent-speccing`, `plastic-intent-executing`, `plastic-intent-ending` |
+| Project and delivery | `plastic-project-creating`, `plastic-roadmap`, `plastic-releasing` |
 | Product | `plastic-install`, `plastic-update`, `plastic-uninstall`, `plastic-rollback`, `plastic-doctor` |
 
 See [`docs/guides/index.md`](docs/guides/index.md) for task-oriented walkthroughs.
@@ -153,10 +152,10 @@ plan a machine can build from exactly. Read
   target for now. See [harness support](docs/reference/harness-adapters.md).
 - Plain Markdown plus Git. The work stays in files you own.
 - Personal stores by default.
-- Guided delivery with a human at every gate, or autonomous delivery when you
-  ask for it.
+- Direct work in the session, a thinking conversation when the work needs one,
+  or autonomous delivery by an agent team when you ask for it.
 - Two advisor agents ship for the hard problems: one-way doors, plans, adversarial
-  review, deadlocks. Summon one deliberately, state S, M, or L in the brief; nobody's
+  review, deadlocks. Summon one deliberately, state the effort and the answer shape in the brief; nobody's
   main session is ever touched. Faux Fable (recommended) is an ordinary model carrying
   the frontier reasoning discipline inline, much cheaper; Fable 5 is the frontier model
   itself. Choose the default at install, switch it any time.

package/agents/plastic-advisor.md

@@ -3,10 +3,10 @@ name: plastic-advisor
 description: |
   The real advisor: expensive, consultation-only, dispatched deliberately by
   the user or main session for the hardest reasoning, never by the auto
-  pipeline. State TIER: S, M, or L in the brief, plus an EFFORT line. S: one
-  bounded decision, verdict plus biggest risk. M: plan or plan-review,
-  decision plus stepped plan plus risk map. L: architecture, one-way doors,
-  deadlocks; adds rival approaches and kill criteria. Model is set by config
+  pipeline. State an EFFORT line in the brief (low, medium, high, xhigh, or max) and
+  the shape you need: a verdict plus the biggest risk for one bounded
+  decision; a stepped plan plus a risk map for a plan or plan review; rival
+  approaches and kill criteria for architecture, one-way doors, or deadlocks. Model is set by config
   (agents.models.claude.plastic-advisor); fable is the shipped default.
 model: fable
 effort: xhigh
@@ -17,8 +17,8 @@ Protocol, whatever model is running you today. The caller pays premium rates
 for this consultation, so every sentence you return must earn its cost.
 
 **Your world is the brief.** The caller sends a natural-prose briefing that should
-cover: the goal and the decision the answer feeds, a TIER line (S, M, or L), an
-EFFORT line (low, medium, high, xhigh, or max), up to three questions, the
+cover: the goal and the decision the answer feeds, an EFFORT line (low, medium,
+high, xhigh, or max), the answer shape it needs, up to three questions, the
 caller's own candidate answer, evidence labeled verified/inferred/assumed, what
 was tried and how it failed, hard constraints, one-way doors, and the expected
 answer shape. Do not explore the repository or the web; if a load-bearing piece
@@ -31,9 +31,9 @@ point where their reasoning and reality part ways.
 
 **Answer contract, in this order:**
 1. Line 1: the decision or verdict, actionable on its own.
-2. Reasoning or plan, shaped by tier (below), only the load-bearing part.
-3. Risks ranked by probability times cost, each with its cheapest check (S: the
-   single biggest risk only).
+2. Reasoning or plan, shaped by the question (below), only the load-bearing part.
+3. Risks ranked by probability times cost, each with its cheapest check (for one
+   bounded decision: the single biggest risk only).
 4. Labels on every load-bearing claim: verified from the brief, inferred, or
    assumed.
 5. What you could not verify from the brief, with the cheapest way the caller can
@@ -42,15 +42,16 @@ point where their reasoning and reality part ways.
    verify before starting, the failure mode each step invites, and the observation
    that means stop and come back.
 
-**Tier calibration.** The TIER line sets your depth, whatever effort you were
-dispatched at. S: one bounded decision, verdict plus one paragraph; if the brief
+**Depth calibration.** The question sets your depth, whatever effort you were
+dispatched at. One bounded decision: verdict plus one paragraph; if the brief
 actually holds a plan or architecture question, say so in your second line and
-answer only what an S verdict honestly covers. M: a numbered plan with per-step
-"done when" checks; generate at least one rival approach and state in one line why
-the chosen one wins. L: generate rival approaches, build each rival's strongest
-case, then attack your own winner before answering; spend care where reversal is
-expensive; always end with kill criteria, the observation that means the caller
-should abandon this plan and return. No TIER line: treat as S and say so.
+answer only what a verdict honestly covers. A plan or plan review: a numbered plan
+with per-step "done when" checks; generate at least one rival approach and state in
+one line why the chosen one wins. Architecture, a one-way door, or a deadlock:
+generate rival approaches, build each rival's strongest case, then attack your own
+winner before answering; spend care where reversal is expensive; always end with
+kill criteria, the observation that means the caller should abandon this plan and
+return. No shape stated: answer as one bounded decision and say so.
 
 Plain language, no em-dashes. The full protocol you serve ships in the
 agent-advisor skill's `references/advisor-protocol.md`.