Recently added
Plugins grouped by the day they were added to the directory, newest first.
Sep 19, 2026
- legacy-plugin58stars
legacy-plugin is an open-source Claude Code plugin indexed from GitHub. Review its manifest and source before installing.
Claude Code - my-otel-plugin58stars
Queries OpenTelemetry traces in Tempo.
Claude CodeAgent Plugins - research-agent0stars
Personal academic research assistant — arXiv/Semantic-Scholar ingestion, topic hierarchy, LLM gap analysis and reports, and paper ratings.
CodexClaude Code1 skill1 MCP server - problem-solving-memory0stars
Shares one authoritative Problem and its typed evidence across supported AI coding hosts.
CodexClaude Code1 MCP server - caliber-jev-compaction1.3kstars
Verbatim Jev-guided compaction for Claude Code sessions.
Claude Code - spoken-tts4stars
Opt-in per-conversation spoken summaries via <spoken> tags, plus a CLI to speak a named passage. Providers: edge-tts (recommended) and macOS say.
CodexClaude Code1 skill - codexbar-quota-handoff4stars
Reminds Codex CLI to wrap up when CodexBar detects the Codex 5-hour or 7-day quota is nearly exhausted, and to ask before writing a handoff document if needed.
CodexClaude Code - cheap-dev-workers4stars
Cheap workers for exploration, verification, and safe-log summaries.
CodexClaude Code - charley-skills4stars
Charley's Agent Skills for engineering, memory, and media.
CodexClaude Code15 skills - clear-ui0stars
A calm, adaptive status bar for Claude Code: model, effort, project and git branch on the left; context, 5-hour and weekly usage on the right. One row when the terminal is wide enough, two when it is not. Drawn only from the data Claude Code already provides plus one cached git status — no network, no credentials, no transcript parsing. Optional companion to clear-claude.
Claude Code3 skills - clear-claude0stars
Make Claude Code easier to understand without making it less capable. Ships the Clear Partner output style — answer-first, plain English, concise by default, deep when asked — plus two read-only diagnostic skills, clear-doctor and clear-audit.
Claude Code2 skills - spike-ui-render0stars
Disposable function-hooks spike. Not shipped, not in the marketplace.
Claude Code - spike-tool-observer0stars
Disposable function-hooks spike. Not shipped, not in the marketplace.
Claude Code - spike-state-compact0stars
Disposable function-hooks spike. Not shipped, not in the marketplace.
Claude Code - spike-hello-runtime0stars
Disposable function-hooks spike. Not shipped, not in the marketplace.
Claude Code - q0stars
An agentic coding workflow framework: skills for planning, implementing, verifying, and grooming, grounded in per-project conventions docs.
Claude Code15 skills - agent-ops-skills1stars
Skills from agent-ops-patterns: hand your agent a source (a course, a paper set) and get a disposition record with a quoted file:line or a scoped command with its exit code and output per verdict. First skill: Stanford CS329A "Self-Improving AI Agents".
Claude Code1 skill - ccc-review0stars
CCC Review (Claude Code ↔ Codex Review): Claude Code independently reviews Codex's changes; findings go back to Codex in a bounded loop.
CodexClaude Code1 skill - deckrd0stars
deckrd: your goal to tasks framework. Includes 2 skills and 3 MCP servers.
Claude Code2 skills3 MCP servers - bdd-coder0stars
BDD implementation skill — strict Red-Green-Refactor cycle for any coding task
Claude Code2 skills - jyotish0stars
Конвейер разбора натальной карты: десять этапов методики, сбор данных с vedic-horo, детерминированные расчёты и контроль качества
Claude Code1 skill - copyeditor0stars
Polish writing through an explicitly configured MCP server.
Codex1 skill1 MCP server - copyeditor0stars
Polish writing through an explicitly configured MCP server.
Claude Code1 skill1 MCP server - mathcast0stars
Turn maths and science learning material - Markdown, PDFs, links - into narrated Manim animations. Ingest to an outline, plan the scenes, write the narration, render, repair what breaks, inspect the frames, and promote to final quality.
Claude Code4 skills - arkitect2stars
Your agent draws the architecture. Creates and edits native, editable Draw.io (.drawio) and Excalidraw (.excalidraw) diagrams - cloud and system architectures, data pipelines, agentic systems, network views and flow diagrams - with 6,000+ bundled icons, a learned house style, and everything running locally.
Claude Code6 skills - scio2stars
Read from and contribute to the encyclopedia written and verified only by AI agents: research with sourced claims, propose articles, sit on review panels, contest decisions — with your rank and permissions checked live. Say /scio:register once: the agent registers itself and the key is saved locally (one agent per model); no environment variable needed.
Claude Code1 skill2 MCP servers - rule-of-two2stars
Two adversarial reviewers from different model families - a Claude agent and a Codex agent - tear apart a Claude Code agent, skill or plugin against one shared rubric and return a viability verdict with cited defects, weaknesses and the edits worth making. Use when reviewing a skill, agent, command or plugin as an artifact: 'is this skill worth shipping', 'tear apart this plugin', 'will this agent's description actually trigger', 'review my SKILL.md', 'does this duplicate something we already have'. Coverage is a reported outcome: if only one reviewer runs, or the two models cannot be resolved to different families, the report says so in its own title rather than passing a single-family opinion off as two. Do NOT use for architecture, implementation plans, product ideas or code diffs - use crew's architect-reviewer, planner or code-reviewer. Standalone - its own two-model config, never crew's. No hooks.
Claude Code - dev-workflow1stars
Nine skills for research, brainstorm, plan, review-plan, execute, review, compound, handoff, and propose — with triage, convention compliance, and knowledge compounding
Claude Code9 skills - bouncer1stars
A decision layer for Claude Code hooks. Classifies proposed tool calls against a policy you own and returns allow / ask / deny, using a fast System One model instead of a frontier LLM. Observe-only until you have looked at your own calibration table.
Claude Code - web-launch9stars
A reusable, stack-agnostic website launch system: a verification-gated checklist (Done ≠ Verified), a refuse-to-launch sign-off gate, a Day 1/3/7/14/30 monitoring cadence, an anonymized failure-mode library, and programmatic-SEO-at-scale quality gates. Routes to specialist SEO/security plugins for commodity crafts.
Claude Code2 skills - monkeys0stars
Spend an API key, token or password from the OS keyring on one command, without the value entering the conversation.
Claude Code1 skill - clarity-reminder0stars
Holds a commit until the pass that applies to it has run.
Claude Code - cmd0stars
Persönliches Planungs-Toolkit: Vorhaben klären (plan-grill), Plan reviewen (plan-review), umsetzen (plan-execute), durch die Projekt-Einrichtung lotsen (project-setup), Projekt-Settings einrichten (project-settings), Projektablage strukturieren (project-structure), den projekteigenen Prompttext prüfen (project-audit), CLAUDE.md härten (project-rules), sie kuratieren (project-curate), aus der Session lernen (session-learn), den Arbeitsstand übergeben (session-handoff) und ihn im neuen Fenster wieder aufnehmen (session-resume).
Claude Code11 skills Coach technique pour personnes non techniques : elles construisent un vrai projet avec l'IA, et apprennent en le construisant.
Claude Code1 skill- zhide915-skills0stars
zhide915's personal skills
Claude Code - agent-factory0stars
Agent roles, house rules, and repo provisioning commands for turning objectives into shipped code.
Claude Code7 skills - hatsu0stars
Hatsu is the local plane of the Akatsuki system — a lead persona, a small roster of focused independents, and the way of working itself, running on your own credentials with no GitHub App and no bot identity. Kurapika (/kurapika) leads, declaring one of six Nen-type work-modes in every reply: Enhancer writes product code; Conjurer authors canon and governance as conjured contracts with conditions; Transmuter shapes machinery; Manipulator drives GitHub-side operations — drives, wakes, labels; Emitter cuts releases and fans out the repin; Specialist takes product intake, his kept Product-Owner canon. Eight independents stand around him, seven ratified and one provisioned: Gon, the mission-scoped trusted delegate who asks what the mission is, which named gates he may cross and under what conditions — and who, until his delegation grammar is ratified, crosses none; Hisoka, UI/UX review and quality measurement before a PR is ever posted; Phinks, adversarial QA under the proven-finding discipline, before a release and, from v0.5.0, on a release-adjacent change set pre-PR; Uvogin, performance testing against the fixed seven metrics with method blocks and baselines; Feitan, security and security only — auth flows, secrets and credential handling, network and storage boundaries, data minimisation — citing SEC-{n} by id, resolved and never remembered; Chrollo, architecture and handbook conformance against the UZF core, exactly one resolved stack handbook and the repository's own architecture notes, who reviews the handbooks and never authors them; Netero, the chairman who observes Hunter execution and files complete issues when constitution, canon or machinery need enhancement; and Illumi, PROVISIONED rather than ratified for En's pre-Ready observation hold and no other loop, read-only, waking Kurapika and acting on nothing. Feitan and Chrollo were activated from the Genei Ryodan bench on 2026-09-09 and their definitions land here at v0.5.0; Killua's role, the rest of Illumi's proposed row and the five remaining bench profiles stay OPEN in docs/ROSTER.md — proposals, not rulings. Hatsu depends hard on the Nen CLI: every deterministic step is a Nen verb, the dependency is checked at warm-up against nen/contract.json (nen's own location and shape, validated by nen schema check), and it fails closed with auto-install — the checksum-verified bootstrap runs itself, and only if that bootstrap fails does the session halt with the exact command. A Nen-owned operation is never improvised in prose. The skill surface is thirty-nine at v0.24.0 (thirty-eight at v0.6.0–v0.7.0, then byakugan). Seventeen answer a request — backlog-state, backlog-board, backlog-loop, backlog-synthesis, bankai-handbooks, bankai-quality, build, file, futon, getsuga, izanagi, izanami, jujisho, pr-state, senkei, sharingan, tensho — each ported under its existing name onto Nen verbs, with drive renamed to sharingan at v0.5.0 (same behaviour; hatsu:drive no longer resolves). Twenty-one ARE the way of working. Ten shipped at v0.4.0: breath warms the effort up, cuts the branch and proves the base tip builds, rasengan AUTHORS the change the request asks for, kokusen verifies the finished tree with the declared iteration checks and refuses a red one before it commits with an ask on every flagged file, amaterasu launches from your own working directory, tsukuyomi runs the tests, rikugan renders the rich HTML report, jutaisho rings the bell, ao pulls from the base, aka pushes when YOU call it, and ren is the per-request loop over them that never pushes. Eight are the PR side, new at v0.5.0: hanten reviews adversarially pre-PR and routes by scope to Hisoka, Feitan, Chrollo, Uvogin or Phinks in one fixed finding shape; gyo is linting; byakugan holds the touched-file coverage ladder and never lowers the bar; kotoamatsukami runs impacted unit, UI and integration suites; shibari composes and opens the one PR; jujutsu pairs a physical device and lands it as a launch target through a PR; murasaki pulls and pushes; mukai is the human-called review-and-PR phase that ends by immediately starting en; and en is the izanagi-capped readiness watch that owns deterministic current-head readiness, with Illumi observing the pre-Ready hold only on surfaces that support an in-session subagent. Three close the release side, new at v0.6.0: susanoo runs the lane's declared archive locally and uploads nothing, kagutsuchi prints a non-production deploy plan always and acts only on your own call naming the target, and mugetsu publishes on your recorded per-target go behind G3 — so four of the five phases only you may call have a skill (aka, mukai, kagutsuchi, mugetsu); the fifth is the merge, which has none because G2 is an action no agent performs. Neither of the last two is ever reached from a composite. Every parameter lives in two files: nen/contract.json's project block says what nen executes, nen/workflow.json says what the workflow decides — branch template, iteration checks, the 80/85/90 coverage ladder, reports, notifications, commit trailers, the monitor cap and the model matrix; docs/WORKFLOW.md is the authority on both. hooks/hooks.json ships two harness hooks: a Stop bell off the gate-stop marker, and a PreToolUse guard that refuses a git commit or git push on the base branch. Five phases are yours alone to call — aka, mukai, the merge, kagutsuchi, mugetsu — and only five G5 conditions ever interrupt you. Two provenance trailers, one per plane, from v0.8.0: Hatsu-Agent is its provenance trailer -- what a local Hatsu session writes, on your own credentials -- and Akatsuki-Agent is the CI plane's, written only by an Akatsuki roster agent there. A local Hatsu actor writes only the first and refuses the second, because a persona is not the CI plane; an autonomous Akatsuki actor writes only the second. nen/workflow.json admits both so one hook can validate either truthful plane, which is not licence to choose one by default. Neither is AI attribution -- each names which agent of which plane did the work -- and no other AI attribution trailer is ever recorded, with includeCoAuthoredBy: false the recommended setting. Existing commits are not rewritten. Per-skill evidence is in docs/ab/. From v0.7.0 Hatsu also runs on two more surfaces: the same skills and personas are generated into Codex and Cursor layouts under surfaces/ by nen surface mirror, placed into the repository you are standing in by the warm-up (.agents/skills/ plus an untracked AGENTS.override.md, or .cursor/skills/ plus .cursor/agents/) and excluded through .git/info/exclude rather than your .gitignore — invoked as $name on Codex and /name on Cursor, hatsu:name on Claude Code. The Codex persona file is an OVERRIDE: it replaces your AGENTS.md in the envelope rather than joining it, so the warm-up copies yours into it verbatim first and never writes the tracked file. Two absences are named rather than assumed: neither of those surfaces has a turn-end hook, so jutaisho rings the bell in-session and says so, and Codex has in-session subagents (spawn_agent); hanten still raises an isolated reviewer as a second codex exec run in its own worktree. The model matrix carries a column per surface — the reviewer tier is deep everywhere, opus / sol / grok — Cursor is Cursor-native only, and the frontier tier never runs a subagent on any surface. docs/SURFACES.md is the authority; scripts/surface_mirror_check.sh fails a mirror that has drifted from its source. At v0.11.0 the pin moves to nen v0.7.0 and nine more residues retire with it. The largest: izanagi's mandatory cap is no longer counted in a skill's prose -- nen loop iterate claims each mutating iteration against the cap its own invocation states and refuses the claim past it, so izanagi and en stop bookkeeping, a cap cannot be widened by re-typing the line, and en's ledger outlives the session so a resumed landing continues the same budget rather than restarting it. stage triage grows the two detectors kokusen, tensho and jujisho were each checking by eye -- local-config (the .local filename infix) and large (--large-bytes, default 1 MiB). pr ready READS nen/gates.json's dependabot_carve_out, inert data until now, and never silently: a satisfied row carries its own note and meta.dependabotCarveOut says whether it fired. Every readiness verdict says which binary decided it, path included, and pr-state, sharingan and shibari relay that line rather than summarising it. Every relative own-path flag resolves against --repo's root instead of the process's directory, which file, build, futon, senkei, shibari and bankai-quality each had to warn about. A missing or malformed --target exits 2 across sixteen verbs and an unreadable input exits 2 on split verify, so an invocation mistake stops reading as a registry finding or an incomplete split. changelog collate's printed manifest agrees with the section it wrote, so getsuga relays it instead of noting a defect; issue attach-sub prints the sub-issues fallback and names nen issue edit-body as the write that performs it; parse izanami accepts a single-quoted --jq on a gh api read; chain-position and effort classify say which of their values can refuse a verdict; and nen bootstrap --help publishes the whole exit-code contract, including the 7 only it can return, while its cache slot is keyed on the source as well as the ref. Four behaviours change in place at this minor with no new flag to notice them by, which is why nen's own compatibility floor is seeded at 0.7 and a pin below it is refused by name. At v0.12.0 the maintainer's ruling of 2026-09-10 re-scopes three of the loop's phases without touching a single authority: rasengan is the AUTHORING phase -- it builds the thing, on the stack nen/contract.json declares, running the iteration checks as the author's own inner-loop feedback rather than as a commit gate; kokusen carries the compile-before-commit, running those same checks over the finished tree and refusing to commit on red with the failing check quoted; and breath's proof is stated for what it always was, a verdict on the BASE tip taken before a line is authored, where a red one is a G5 stop and never this effort's to repair. ren's order is six whole steps again -- breath, rasengan, kokusen, amaterasu, rikugan, jutaisho -- with the interim half-numbered work step folded into rasengan where it belonged, and five load-bearing relations stated: no authoring on an unverified base, nothing committed that was not authored in this turn and verified in it, the launch shows the committed tree, the report quotes the launch that ran, the bell carries the report. The five G5 conditions that interrupt a running loop are unchanged. At v0.13.0 the range stops being computed in this plugin's prose, because nen now decides it: the maintainer's ruling of 2026-09-10 is "exact minor is fine, unless there is a breaking change", and nen v0.8.0 ships COMPATIBLE_MINOR_FLOOR -- the lowest dependency.minimum pin a build satisfies, printed as a 'compat floor:' line on every nen shu tools run and carried in --json as compatibleMinorFloor. The warm-up probes nen --version for presence only and then reads the verdict off the nen row of nen shu tools --repo <the Hatsu checkout>: a satisfied row proceeds to the report, a pin below the build's floor is refused by name with the repin stated, a binary older than the pin re-pins through nen bootstrap, and an absent verb means a binary older than anything this plugin supports. The report line carries the floor beside the version, and says 'floor not reported' on a binary older than 0.8.0 rather than inferring one. dependency.minimum stays "0.7" while pinned_ref moves on its own to v0.8.0 -- two values that no longer travel together, since minimum moves only when nen's CHANGELOG carries a real bullet under Breaking / consumer notes. Left as it was, every warm-up on a host carrying nen 0.8.0 would have read out of range and rebound ~/.local/bin/nen down to v0.7.0, making the floor inert for every Hatsu consumer. Pinned to nen v0.8.0, minimum 0.7. At v0.13.1 the remaining echoes of the literal `nen 0.7.0` outside the dated docs/ab/ transcripts and the version-history prose above are repointed to name the pin by reference -- "the pinned build", or the contract's own pinned_ref -- because a convenience copy that still read v0.7.0 once the contract moved to v0.8.0 was exactly the drift nen/contract.json's own authority clause calls a bug (zheref/hatsu#43). Both generated surfaces are regenerated against nen 0.8.0 to match. At v0.14.0 every path a skill builds from the plugin root is spelled $hatsu_root and resolved in the very shell that uses it: hatsu-warmup's section 0 resolves the root, prints it as one quoted line and reads the contract in one shell; every later block either carries that resolver or sets the variable from the printed line; the manifest is read structurally as the one shape Claude Code's tooling writes, every line validated and any other shape refused; and the captured path is proved to be the candidate's directory, refused if it holds a newline, and taken with CDPATH cleared. So the mirrored skills resolve the plugin root on Codex and Cursor, where $CLAUDE_PLUGIN_ROOT is unset or names another plugin, with the same lines that run on Claude Code (zheref/hatsu#38). At v0.18.0 Hatsu adopts Nen v0.10.0 and its explicit review-round-only policy: Copilot supplies the sole automated round, no separate APPROVED review is required, and the maintainer keeps the human merge decision. At v0.19.0 Mukai ends after opening the PR, rendering its landing report, and immediately starting En; En owns deterministic current-head readiness and uses paced pre-Ready observation without spending cycles on quiet reads, and the human merge remains outside the run. At v0.20.0 aka no longer runs project-wide tests: tsukuyomi is focused tests on every ren turn, kotoamatsukami owns impacted regression at mukai immediately before gyo extracts coverage, and aka is lint/squash/ao/push. At v0.21.0 every Rikugan page opens with This last turn (mukai adds corrections, local-check fallout, and half-run stops) and Architecture delta is a change-highlighted structural diagram rather than a file-line inventory. At v0.22.0 00 stays last-turn only while 01–07 cover the whole session against the base, on every process or product effort. At v0.23.0 gyo is linting on every Ren turn and kotoamatsukami owns coverage capture, extraction and gating at mukai. At v0.24.0 the skill surface is thirty-nine: byakugan captures and measures coverage independently of tests; kotoamatsukami owns unit, UI and integration suites only; gyo remains lint. At v0.25.0 Netero is ratified as process chairman; Breath cuts a fresh branch for every new effort rather than reusing a clean feature branch implicitly; G5 stop reports carry blocker evidence on the turn variant. At v0.26.0 Hanten records per-effort reviewer budgets in `.nen/hanten/<branch-slug>.cycle.json` so remediation cannot re-raise Chrollo twice or Hisoka and Uvogin past their maxima. At v0.27.0 the skill surface is forty: third-hand lands as the wrap-up harvest after En. At v0.28.0 it is a separate phase that starts once En has completed, not En step 8; Codex asks through request_user_input and raises Netero in-session (spawn_agent); Antigravity asks through ask_question (is_multi_select) and raises Netero with invoke_subagent Workspace inherit. The merge stays a human G2 act with no skill. At v0.29.0 Illumi's En watch uses Codex in-session spawn; generated surface inventories are 41 skill files (forty plus warmup) and 9 personas. At v0.30.0 the Hanten cycle ledger is fail-closed: Breath inits it after the branch cut, decide/record/show refuse a missing file, load-mutate-save is locked with a unique temp path, and skipped-exhausted is refused while used is under max. At v0.31.0 the warm-up keeps a consumer plugin checkout current before it refreshes a target: scripts/hatsu_plugin_update.sh --auto fast-forwards trunk or the newest vX.Y.Z tag, skips dirty and authoring trees, never discards, and on Claude Code runs claude plugin update hatsu@hatsu; docs/SURFACES.md is the authority on pointing every surface at a local checkout. Current dependency: nen v0.10.0 with minimum 0.10, required for review-round-only approval policy; Nen compatibility floor 0.7 remains unchanged. At v0.33.0 every skill becomes reachable from a cold checkout: docs/STANDALONE-ENTRY.md is the new authority, carrying the five-step preamble (warm up, orient, establish the delta, elicit, declare) and the four state classes a phase used to inherit from its caller -- the turn boundary, a mandatory argument, a position in an order, and a predecessor's artifact. Twenty skills gain a ## 0. Standalone entry section -- thirteen that derive state cold, seven that add the warm-up, orientation and expectations only -- and the delta default everywhere is the FETCHED origin/<branch.base>, union of committed and uncommitted; the base is the LATEST trunk by the maintainer's ruling of 2026-09-18 -- git fetch origin first, then origin/<branch.base>, with the local ref used only where the fetch has just proved it equal -- and against <base> is the declared delta clause on byakugan, kokusen, tsukuyomi, kotoamatsukami, rikugan and hanten, verified live at nen 0.10.0 including the two-clause and enum-plus-base shapes, while ao and murasaki keep from <base> because a delta is read against a reference and a catch-up pulls from a source. A fetch that fails, an absent origin, or an unresolvable base stops and names the condition rather than falling back to a ref that cannot be known to be latest. Breath runs at most once per session and is never skipped when owed, and it now guarantees an OUTCOME rather than only a verdict: however the checkout looked, a standalone run ends as a branch off a main that was PROVEN green, carrying the maintainer's own commits on that proven tip, with their uncommitted changes still present and still uncommitted. Four steps -- preserve (git stash push --include-untracked, the SHA captured and printed, addressed by SHA and never stash@{0}), prove (every iteration.checks entry against origin/<branch.base> in an isolated worktree holding none of the effort), place (the commits replayed onto the proven tip through ao, which rebases when nothing is published and MERGES when something is, because a published commit is never rewritten), restore (the stash reapplied, still uncommitted, and never dropped on a failure). The win is that a red after the restore is the effort's, mechanically rather than by argument, so a red base is a G5 taken before anything returns and the stash is restored before that stop. Rewriting a commit is confirmed first; stashing is not, being reversible. Hanten routes a missing ledger to breath rather than asking the same question twice. docs/STANDALONE-ENTRY.md joins the plugin-bump guard's surface and the shared-policy preamble, so a runtime-read authority cannot ship edits to nobody. Breath inverts its new-effort default when typed by hand, so work in progress on the current branch reads as the continuation it is; tsukuyomi derives its lane from the delta and offers to have missing focused tests authored rather than refusing; hanten enforces breath when the cycle ledger is absent and asks for scope rather than raising the whole bench; sharingan resolves an omitted PR number from the branch; rikugan recovers turns from the session or from commit history and marks what it cannot recover; third-hand resolves this sitting from the branch or an open PR. kagutsuchi and mugetsu gain orientation and no authority whatsoever -- G3 is unmoved, the invocation is never the go. The second rule is the load-bearing one: no skill is ever indefinitely independent. A standalone run absorbs no later phase, creates no standing authority, crosses no gate the wired run would not cross, and ends with a hand-back line naming its successor; ren, mukai and en state that the phases they call skip ## 0 entirely. murasaki typed alone does its own step 1 and finishes it -- fetch, catch the branch up with origin/<branch.base> through ao, resolve the mechanical conflicts, stop at G5 on a semantic one -- declining only what was never its; rasengan alone stays caller-bound, because the request is its input and no checkout supplies one. At v0.36.0 a tag records a build that LANDED somewhere, by the maintainer's ruling of 2026-09-18 as corrected on 2026-09-19. kagutsuchi section 4a cuts one tag after a --run send that returned exit 0, composing dist/<target>/ from the target it was called with over the IDENTITY in nen/workflow.json's tags.identity; susanoo section 5a reads that same file and merely NAMES the identity the coming tag will carry, cutting nothing and pushing nothing -- one name source, so there is no second name to disagree with. tags.deploy.<target> is read before anything is cut, so a repository declaring only tags.identity gets no tag and an undeclared target is reported rather than tagged; the target is data, quoted and passed as an argument, and the prefix is composed from it rather than pattern-checked. The identity file's second line is the commit the archive was built from, and the tag is cut THERE rather than at HEAD, because the announce and the cut are separate invocations and HEAD can move between them, so its standing property that nothing leaves that machine is literal again. The symmetry is the maintainer's: an upload that succeeds becomes a tag, and a release Apple approves becomes a GitHub release. Both blocks are opt-in and off by default, so a repository declaring neither behaves exactly as before. The NAME is the repository's, read from a declared nameFrom as data -- never templated into shell source, symlinks and out-of-tree paths refused, validated with git check-ref-format before anything is spawned. nen tag cut's refusals stand, including that --at must be an ancestor of origin/<trunk> -- a rule about the COMMIT rather than the branch, so a feature branch at the trunk's tip tags fine; what keeps a tag from attesting the wrong bytes is a separate clean-tree condition, and --trunk comes from branch.base rather than a default. The name carries a dist/<target>/ species prefix because it shares one namespace on origin with getsuga's release tag, which mugetsu proves by name resolution; mugetsu section 3 now says in rule that only the release tag satisfies it. --push is not atomic, and the one sanctioned remedy for a rejected push is deleting and re-cutting a local tag verified absent from origin. mugetsu is still G3 and still cuts no tag.
Claude Code - agent-skills0stars
Personal Claude Code skills. Review: self-review (own branch, pre-PR), guided-review (PR walkthrough, never posts), adversarial-review (skeptical pass, one PR comment), pr-review (rubric pass, inline PR comments). Plus mutation-coverage (mutation testing + test-gap closing), refactor-component (structural refactor of component sources), comment-cleanup (deletes and shortens comments in a diff, a file or a package), pr-description (writes the PR body from the branch diff) and authoring-skills (skill authoring conventions).
Claude Code7 skills - georgeshani-skills2stars
George Shanidze's agent skills — a growing collection covering software engineering, mobile release engineering, and app store publishing. See the individual skill descriptions for what's actually in here today.
Claude Code3 skills - task-poke0stars
Submits a continue prompt when a main-loop turn ends while the TodoWrite or Task tools list still has unfinished tasks, at most 5 times in a row. Turns the task tools on for every model unless CLAUDE_CODE_ENABLE_TODO_TOOLS is set. Needs function hooks (early access).
Claude Code - prompt-time0stars
Draws the time under each of your messages and each text block of the model's replies in the transcript, also for a resumed session. Display only: nothing reaches the model. Needs function hooks (early access).
Claude Code - memory-save0stars
Loads MEMORY.md into the session and saves project learnings to it in the background after every turn, without blocking the stop. Needs function hooks (early access).
Claude Code - limit-watch0stars
Shows the subscription usage limits under the prompt with a reset countdown and a projection, opens a /limits pane, and logs once when a limit passes 80% and 95%. Needs function hooks (early access).
Claude Code - gemini-review0stars
Gemini reviews each git commit the model runs, with the staged diff and the conversation, and stops a commit with a serious bug, a security hole or a committed secret. /gemini-review turns it on or off; gemini-core holds the key, tier, model and thinking level. Needs function hooks (early access).
Claude Code - gemini-plan-review0stars
Gemini reviews each plan before the approval dialog, with the conversation, and sends a plan with a blocking finding back to the model, at most twice. /gemini-plan-review turns it on or off; gemini-core holds the key, tier, model and thinking level. Needs function hooks (early access).
Claude Code - gemini-core0stars
One place for the Gemini key, tier, model and thinking level of every Gemini mod: adds $.gemini, which builds each mod's request and reads Gemini's answer. /gemini-core sets the tier and, per mod, the model (picked from Google's model list) and the thinking level. Several keys are tried in turn. Needs function hooks (early access).
Claude Code - gemini-compact0stars
Moves compaction to Gemini: in summary mode Gemini summarizes the conversation and the newest messages stay; in prune mode every message stays and Gemini keeps, truncates or drops old tool calls. /gemini-compact sets the mode, the threshold and on/off; gemini-core holds the key, tier, model and thinking level. Needs function hooks (early access).
Claude Code - gemini-advisor0stars
A Gemini advisor the model calls by itself: Gemini reads the conversation and the model's message and answers with a second opinion. /gemini-advisor turns it on or off; gemini-core holds the key, tier, model and thinking level. Needs function hooks (early access).
Claude Code - flaky-memory0stars
Remembers which tests failed on which code and tells the model when a failing test also passed on the same code, so it treats a flaky test as flaky. Needs function hooks (early access).
Claude Code - disk-janitor0stars
Measures the build artifacts of the project, shows them on the status line above 5 GB, and deletes the ones you pick in the /janitor pane; data directories are never listed. Needs function hooks (early access).
Claude Code