Adds an `agents/openai.yaml` beside each `SKILL.md` so the skills work in
Codex as well as Claude Code, without generated copies:
- `interface.display_name` + `interface.short_description` for the Codex
skill picker, hand-written for all 39 skills.
- `policy.allow_implicit_invocation: false` on the 22 user-invoked skills —
the Codex analog of `disable-model-invocation: true`, so Codex excludes
them from implicit invocation while explicit `$skill` still works.
- Document the dual-harness invocation model in `.agents/invocation.md`,
`CLAUDE.md`, and the promoted-bucket READMEs.
- Add `AGENTS.md` as a symlink to `CLAUDE.md` so Codex reads the same
instructions; note Codex as a `link-skills.sh` install target.
Slimmed-down rework of the approach prototyped in #522: keeps the essential
cross-harness metadata, drops the Ruby validator, the runtime-detector test,
per-skill `default_prompt`s, and the unrelated promotion changes.
Co-authored-by: gabimoncha <gabimoncha@users.noreply.github.com>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Close a standing CLAUDE.md-invariant gap where two promoted skills were
not fully wired into the plugin manifest and READMEs.
- resolving-merge-conflicts (model-invoked): add to .claude-plugin/plugin.json,
top-level README.md, and skills/engineering/README.md.
- implement (user-invoked): add the missing line to skills/engineering/README.md
(already present in plugin.json and the top README).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Align the remaining callers and docs pages with grilling's body, which now
says "decision tree".
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Drop the software-plan scoping from grilling's description and body so the
interview reads as a stress-test of any plan, decision, or idea:
"this plan" → "this", "enact the plan" → "act on it", "codebase" →
"environment". The technique is unchanged, and grill-me is left as-is.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Research stays a real ticket type — it's a shared blocker downstream
decisions hang on, which the frontier's blocking edges exist to render.
What changes is resolution: because research is AFK, charting doesn't
stop and read. After creating the tickets, the charting session fires a
/research subagent for each research ticket to burn it down in parallel,
capturing findings on a throwaway research/<name> branch with a context
pointer.
Charting's "fire the research subagents" step owns how research is
handled; the Ticket Types bullet just names the resolver, and the
one-ticket-per-session rule carries the research exception.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Before scanning, decide where to look. If the user named a direction, take
it; otherwise read the last ~20 commit messages to bias exploration toward
actively-developed paths. A deepening opportunity in code nobody touches is a
refactor you'll never cash in — leverage only pays off where you keep editing.
Syncs the docs page and adds a changeset.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
People kept reading a wayfinder ticket as an ordinary implementation
ticket — a slice of a build to execute — when wayfinder uses them as
decision tickets: questions whose resolution is a decision.
Introduce "decision ticket" as a light leading word: the skill
description and its opening line name it (and say what makes it one),
and the ask-matt / engineering README blurbs and the docs page match,
while "ticket" stays the everyday word once the term is established.
Record "Decision ticket" in CONTEXT.md so the "avoid: ticket" domain
rule no longer contradicts wayfinder's deliberate use of the word.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Wayfinder is the heaviest, most cognitively demanding flow, and the
wiki shows people misusing it two ways: over-reaching for it on
well-scoped work (it's slower and denser than a single grill), and
losing the way at the handoff after the map (looping straight into
/implement and skipping /to-spec). Sharpen the router's wayfinder
on-ramp to steer both.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a Claude Code plugin marketplace so the promoted skills can be
installed as a managed, read-only bundle (subscribe rather than fork),
alongside the existing skills.sh installer.
- Enrich .claude-plugin/plugin.json with marketplace metadata (version,
description, author, license, keywords); keep the curated promoted-only
skills array.
- Add .claude-plugin/marketplace.json making the repo its own
single-plugin marketplace (mattpocock-skills@mattpocock).
- README: document /plugin install alongside skills.sh, and the
subscribe-vs-fork tradeoff.
- CLAUDE.md: extend the promoted-set invariant to cover marketplace.json
and plugin.json/package.json version sync.
- ADR 0002: record why Claude ships now and a native Codex plugin is
deferred (Codex skills field is single-path + drops symlinks, which
can't express a curated subset of a bucketed repo without a restructure).
Verified: 'claude plugin validate . --strict' passes, and marketplace
add -> install resolves all 21 promoted skills at v1.2.0.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Rule 4 is about skipping polish, not the prototype's fate; the trailing
"then let it go" was a leftover from the old delete-it framing and now
misreads as discard, contradicting capture-as-primary-source. Cut it,
and rename LOGIC step 7 off "let go".
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The depth-based rules already treat any subfolder as private, so the
config needs no change. Document lib/ (implementation) and tests/ as the
conventional two-folder shape, with entry points at the root — a fixed
convention that never needs the config extended for a new folder.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Rename skill to setup-ts-deep-modules (it's TypeScript-specific).
- A package's public surface is now its ENTRY POINTS — every file at the
package root — not one designated index.ts. Implementation lives in
subfolders and is private. This is depth-based and extension-agnostic.
- Packages can expose several small entry points instead of funnelling
everything through one giant barrel index; barrels are discouraged, and
the generated repo docs (step 7) say so explicitly.
- Rules renamed accordingly: entrypoint-boundary-from-app,
entrypoint-boundary-across-packages, tests-through-entrypoints.
- Example package now hides impl in a lib/ subfolder.
Re-validated against dependency-cruiser 16: multiple entry points and
cross-package entry imports pass; deep imports into subfolders fail with
the expected rule from app code, other packages, and tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Step 7 now writes <packages-root>/README.md next to the packages it
governs, and adds a context pointer from CLAUDE.md/AGENTS.md so an agent
discovers the boundary rule.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The install location isn't known at runtime, so a relative path breaks.
Use the literal skill invocation instead.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The "keep it as a primary source" section restated rule 6 and was
padded with no-ops. Fold the one extra bit (capture the answer) into
rule 6 and delete the section: capture the prototype as a primary
source on a throwaway branch, pointer on the issue.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A user-invoked setup skill that wires dependency-cruiser into a
TypeScript repo so each package under the packages root is a deep
module: everything hidden behind its `index.ts` interface.
Ships a `.dependency-cruiser.cjs` template with four error-level rules —
index boundary from app code, index boundary across packages (with
intra-package freedom via $1 back-references), tests-through-the-index,
and no-circular — plus a commented layering stub. The skill installs the
tool, folds a `lint:boundaries` script into the repo's umbrella check,
scaffolds an example package, and self-verifies that a deep import fails.
Rules validated against a live dependency-cruiser 16 run.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The final line of to-tickets/SKILL.md carried a stray `</content>`
closing tag with no matching opening tag — a packaging/serialization
artifact from the planning-skills refactor (386d4ff), not valid skill
content. The same stray tag had leaked into the generated landing
docs for to-tickets and to-spec. Strip it from all three.
Fixes#495
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Reframe the artifact handling as capturing the prototype somewhere out
of main — a throwaway branch — with a context pointer to it left on the
implementation issue, instead of "disposing" of it. The main branch
still keeps only the validated decision.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Rework /setup-matt-pocock-skills to ask fewer, friendlier questions and
align the local-markdown issue tracker with the current spec.
- Triage labels: asked only when the `triage` skill is installed, and then
as a single recommended-yes question ("keep the default triage labels?")
instead of an override interrogation. When triage isn't installed, the
section and docs/agents/triage-labels.md are skipped.
- External PRs as a request surface: no longer a setup question. The
GitHub/GitLab templates still carry the flag, defaulted off; flip it in
docs/agents/issue-tracker.md later.
- Domain docs: default to single-context without asking; multi-context is
only offered when the repo shows monorepo signals.
- Local-markdown tickets: one file per ticket under
.scratch/<feature>/issues/<NN>-<slug>.md, never a single combined
tickets.md. /to-tickets, ask-matt, and the local template now agree, and
the spec file is spec.md (not PRD.md) to match /to-spec.
- Re-sync docs pages for setup-matt-pocock-skills and to-tickets; changeset.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Drop the secondary-source label — the actionable distinction is just
that the prototype is a primary source to retain on a throwaway branch.
Collapse the re-descriptions in LOGIC/UI so they point to SKILL for the
disposal concept and add only branch-specific nuance.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Retain the prototype as a primary source on a throwaway branch linked
from the relevant issue, rather than deleting it, while the distilled
answer remains the secondary source. Keeps prototype scaffolding out of
the main branch without destroying the runnable evidence.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
implement is the central Main Flow step but was missing from the
public Reference list. Add it after to-tickets.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The grill-led idea→ship chain stays the documented front door; wayfinder
is a situational on-ramp (greenfield / huge feature build), reserving
"new default spine" for a potential v2.
- ask-matt router: name wayfinder's concrete triggers
- grill-me / grill-with-docs: signpost up to wayfinder as the
too-big-for-one-session on-ramp
Resolves the main-flow decision on the Skills v1.1 wayfinder map (#120).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Graduate `wayfinder` out of in-progress into the engineering bucket
(plugin.json, top-level + Engineering READMEs under User-invoked, docs
page, ask-matt on-ramp, disable-model-invocation).
Consolidate the 22 pending changesets into 10 coherent v1.1 entries:
one `wayfinder` (rename from decision-mapping + full reframe +
graduation), one `grilling`, one `code-review`, one `tdd`; fold the
superseded to-issues changeset into unify-to-tickets and drop its stray
`</content>` tag. Highest bump is minor → 1.0.1 becomes 1.1.0.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
It documented what the skill does rather than steering the agent — the
publish mechanics it described are already Step 5's instructions, and the
conceptual framing already lives in ask-matt and the docs page.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Executes the `unify-to-tickets` decision (personal-wiki #23), reworking
this PR from "add /to-plan" into the full planning-skills unification.
- Rename /to-prd → /to-spec. "spec" is now the single through-line term
(the skill still opens with "you may know this document as a PRD" for
discoverability).
- Merge /to-plan + /to-issues into one /to-tickets skill; delete
/to-issues. /to-tickets emits a set of tickets, each declaring its
blocking edges — one medium-agnostic artifact that reads as a
sequential plan (local `tickets.md`, edges as text) or a parallelizable
DAG (real tracker, native blocking links). The edges live in the ticket
either way; the medium only decides whether anything acts on them in
parallel.
- ask-matt now routes idea → /to-spec → /to-tickets → /implement.
- Sweep every cross-reference (docs pages, chain diagrams, CONTEXT.md,
ADRs, .agents, setup templates) to the new names and vocabulary.
- Rewrite the changeset to describe the unification.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Fold the redundant 'thin vertical slice cuts through all layers, NOT
horizontal' intro sentence into vertical-slice-rules bullet 1 (it was
duplicating the bullet), and cut to-plan's 'no parallelism / top-to-bottom'
restatement — 'sequential' is already anchored as the leading word. Keeps
the shared core identical across the two sibling skills.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a user-invoked `to-plan` skill — the sequential, HITL sibling of
`to-issues`. Both slice work into tracer-bullet vertical slices, but
`to-plan` produces one ordered sequence you drive by hand, one phase per
fresh context, rather than parallel independently-grabbable issues.
It publishes to the tracker `/setup-matt-pocock-skills` configured, shaped
to it: a single sequential `plan.md` for a local tracker, or a parent
issue with ordered sub-issues (native sub-issues where supported) for a
real one.
Registers the skill everywhere CLAUDE.md requires (plugin.json, top-level
and bucket READMEs, docs page), updates the `ask-matt` router's main-flow
fork to offer both to-issues and to-plan, and adds a changeset.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Wayfinder pinned the literal docs/agents/issue-tracker.md, breaking the
indirection to-issues/to-prd/triage rely on: they resolve the tracker
through the ### Issue tracker block setup writes into CLAUDE.md / AGENTS.md.
In repos that keep agent docs elsewhere, wayfinder silently fell back to
local-markdown. It now follows the same pointer and reads the doc's
"Wayfinding operations" section by name.
Closes#471
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Revert the description change and keep the wide-refactor material in
one place. Reshape the skill into a lean Process (steps) and a
Reference section holding the vertical-slice rules, the wide-refactor
expand–contract rules, and the issue body template as co-located
blocks the steps point to.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a wide-refactor exception to the tracer-bullet rule: a single
mechanical change whose blast radius fans across the codebase can't
land as one green vertical slice. Seed "wide refactor" / "blast radius"
across the description, the explore step (detection), and the drafting
step (strategy), and give the expand–contract sequence in prose.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
