* software-factory: add glossary skill for designing a factory
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* software-factory: add example runs grouped by problem
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* software-factory: separate automated review from human review, add ratchet example
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* software-factory: drop review definitions, keep the stricter wording
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* software-factory: optional observability platform, slot surveys
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* software-factory: slot surveys orient, the agent researches current options
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* software-factory: add Task; example runs become example tasks
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* software-factory: start slow, accelerators and brakes
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* software-factory: start slow with throttled automated triggers
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* software-factory: tighten start slow
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* software-factory: define the three brakes
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* software-factory: tasks work on tickets, PRs or the codebase
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* chief-of-staff: shared wiki first; retro and implement-spec as procedures
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* chief-of-staff: procedure paths are relative to the skill directory
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* chief-of-staff: retro every large subagent run and act on the findings
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* chief-of-staff: retro a one-in-five sample of subagent sessions
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* implement-spec: a fresh implementer subagent per ticket
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* implement-spec: a fresh merger subagent per merge
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* implement-spec: merger subagent need not be fresh
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Tell skills to start subagents with a fresh context window
Some Codex models fork the parent conversation when spawning a subagent,
which inherits its context and defeats the point of a subagent.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Trim fresh-context line to one plain sentence
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Add fresh-context subagent rule to repo standards
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Add issue forms (bug, change to an existing skill) with a one-problem and
duplicate-search check, turn off blank issues, and send questions, new-skill
ideas and other harness or tracker requests to Discussions. Add a short
CONTRIBUTING.md with the rules for AI-written issues.
Tell the GitHub issue tracker template to file only into the user's own
repo, never into mattpocock/skills, which is how #541 to #550 landed here.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Wire up a remote before writing a hosted tracker (#1201), offer to git mv
legacy CONTEXT.md (#1176), comment-then-close (#873), read PR body (#1138),
Blocking in Conventions with native gh flags (#855), name the open-blocker
read (#1118), GitLab -O json and frontier command (#1143).
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* docs(readme): per-agent install instructions
Copilot and Codex install the plugin via the repo's own marketplace,
Gemini CLI via two --path installs, and Cursor, OpenCode, Devin,
Windsurf, Amp and pi via skills.sh with -a. Canonical block and
ADR 0002 updated to match.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(readme): lead with self-updating installs
Claude Code, Codex and Copilot install from the repo's own marketplace,
with each agent's one-time auto-update opt-in next to the command.
Gemini and skills.sh are labelled as updated by hand.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(readme): tighten install section, install block and ADR 0002 update
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs: Claude Code installs from claude-plugins-official in one command
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs: unslop the install section, install block, ADR 0002 update and CLAUDE.md
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* tdd: note what each proposed seam catches and misses
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Add changeset for tdd seam trade-offs
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* teach: anchor workspace paths to the workspace root
SKILL.md used `./` both for its bundled format docs and for workspace
paths (lessons/, reference/, learning-records/, assets/), so agents
resolved the workspace paths against the skill folder and wrote courses
into ~/.claude/skills. Name the workspace root once, drop the bare `./`
from workspace paths, and re-sync the docs page.
Closes#377.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(teach): trim #377 fix to a one-line workspace-root rule
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* teach: vary the correct quiz answer's position
Closes#611.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* wayfinder: no triage labels, no placeholder refs, no research PRs
Three things wayfinder published to the tracker went wrong in real
sessions:
- Maps and tickets picked up a triage-state label (ready-for-agent),
which reads as "ready to implement" (#518).
- Drafted tickets numbered T1-T8 were cross-referenced as #1-#8 in
published bodies, auto-linking to old unrelated issues (#507).
- Research subagents opened a draft PR from their throwaway
research/<name> branch (#576).
Say each rule where the step happens. The research docs page drops the
#576 known-problem note.
Closes#518Closes#507Closes#576
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* changeset: credit the wayfinder issue reporters
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* wayfinder: tighten #518/#507/#576 fix wording
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* wayfinder: resolve each ticket as its label's type (#625)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
implement still said "use /tdd" and "use /code-review", the bare
/skill-style prose that #878 replaced everywhere else with an explicit
Skill tool call (the convention in .agents/invocation.md). Both targets
are model-invoked, so the call is allowed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Interpolating the summary into a double-quoted argument let the shell execute backticks and $(...) and expand $VAR, silently truncating the background agent's prompt. Write the summary to a temp file and pass it after -- as "$(cat <file>)", which the shell does not re-parse.
Closes#562.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
to-tickets said only "use the native sub-issue relationship", and the
GitHub tracker template never gave a command for it, so tickets cut from
a spec issue landed as top-level issues. The template now carries a
concrete sub-issue operation (gh issue create --parent / gh issue edit
--add-sub-issue, with a gh api fallback), referenced by wayfinder's child
ticket step; GitLab gets its Part-of equivalent. to-tickets attaches
each ticket to its source issue.
The issue template's "Blocked by" section is now omitted when blockers
were set as native edges, so they are no longer recorded twice (raised
on #262).
Fixes#554.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
SKILL.md only said to save to the OS temp directory, so agents took several attempts on Windows and wrote to a different place on each run on macOS. Reuse improve-codebase-architecture's wording ($TMPDIR, falling back to /tmp or %TEMP%) and tell the user the absolute path. Re-sync the docs page.
Closes#272.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The GitHub tracker template asked `gh pr list --json` for
authorAssociation, which gh does not expose, so the command always
failed. List open PRs with `gh api repos/{owner}/{repo}/pulls` instead
and drop OWNER/MEMBER/COLLABORATOR rather than allow-listing
first-contribution values, so FIRST_TIMER authors are kept.
Fixes#468.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- SCOPE.md: the two-part bar (observed failure + fits the philosophy),
frozen misc/, no new skills, links to every .out-of-scope file
- .out-of-scope: native-question-tool, harness-name-collisions,
new-skills, frozen-misc-skills; tracker rule now names GitHub, GitLab
and local markdown explicitly and rejects Jira (#258, #1137)
- docs/agents/triage-labels.md + Agent skills block in CLAUDE.md
- .github/ISSUE_TEMPLATE: bug and idea forms, blank issues off,
questions routed to Discussions
- README + install block: plugin updates lag the official marketplace
pin; add a "stuck on an old version?" escape hatch
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
New issues get needs-triage. Issues labelled needs-info close after
14 days with no reply; a reply puts them back in needs-triage.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
