sshlg-skills 1.69.0 → 1.71.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md CHANGED
@@ -1,3 +1,65 @@
1
+ ## v1.71.0 (2026-10-20): ten members re-trimmed, the listing fits again
2
+
3
+ - **Pins:** make-skill 0.33.0 (a description is a router: target 400, warning above 500,
4
+ failure only above 1024), agent-sync 1.23.0 (SessionStart prints one line or nothing; the
5
+ guard waits at most 1.5 s for the lease remote), and the description trims of sheleg-dev
6
+ 0.15.0, telegram-dev 0.4.0, xr-dev 0.4.0, web3d-dev 0.2.0, super-ux 0.63.0, sheleg-design
7
+ 1.66.0, seo-aeo-audit 0.30.0 and agent-stack 0.30.0. Each moved the trigger phrases it cut
8
+ from a description, verbatim, into a "Triggers" section of the same SKILL.md.
9
+ - **Family description total 32,495 to 17,601 characters** over 39 skills
10
+ (`description-total`). On the operator machine the skill listing had hit its 30,000 budget
11
+ and shown 81 of 134 skills as a bare name.
12
+ - **The routing hook's words are checked against the whole SKILL.md**, not the description
13
+ (`test/triggers_test.js`, `test/advertised_check.js`): every one of the 11 members still
14
+ carries every routed trigger.
15
+ - **The description reserve is measured against 500**, make-skill's new soft cap
16
+ (`test/validate.py`, `test/route_coverage.js`, `test/audit_bundle.py`); the ratchet stands
17
+ at 17 until task-pipeline's own trim lands.
18
+
19
+ ## v1.70.1 (2026-10-20): context diet, quieter hooks, one source for operator rules
20
+
21
+ **v1.70.0 is a burned tag and published nothing.** It was pushed onto `cb82183` (the 1.69.0
22
+ tree) while the merge it was meant for was still blocked by the required `validate` check;
23
+ its release run 39370618429 was cancelled before publishing, tags here cannot be deleted, so
24
+ the release is 1.70.1. Cause: the tag was cut from `origin/main` without first confirming that
25
+ main had moved to the merge commit.
26
+
27
+ Measured on the operator machine before the change: the global `CLAUDE.md` was 50.6 KB
28
+ (~12.4k tokens, 741 lines), the skill listing hit its 30k budget with 81 of 134 skills shown
29
+ without a description, and a real session start injected about 2.4k characters of hook text.
30
+
31
+ - **Routing block, smaller.** The map region no longer carries the member roster table: its
32
+ "what it closes" column repeated the router table's "answers" cell, and entry points are
33
+ reached through the catalogue search the protocol mandates. The protocol keeps every rule
34
+ (search first, print the plan, list skills used, signature, star, refusal semantics,
35
+ boundary, the no-em-dash rule) in about half the text. The Fabric router lost its Linux and
36
+ Windows paths (routing-registry-total 11902 to 11750). The whole block measures 16,116
37
+ characters, down from 18,395; its budget in `test/router_compaction_test.js` drops to 16,200. Machines with no rendered record
38
+ refresh the 1.69.0 wording instead of keeping it as a hand edit (`PROTOCOL_1_69`,
39
+ `MAP_1_69`; two fixtures in `test/protocol_test.js`).
40
+ - **Hooks say less, and less often.** The SessionStart pointer is one sentence shorter by half
41
+ (442 to about 230 characters) and is skipped on `compact` and `resume`, where the global
42
+ instructions are reloaded anyway. The delivery route is no longer named for a reaction
43
+ («спасибо, отличная фича»), a question with an advice marker («внедрить или нет? дай
44
+ совет») or an audit that forbids edits («сделай аудит хуков, ничего не меняй»): seven
45
+ prompts from the hook audit are silence fixtures in `test/route_coverage.js`.
46
+ - **A gate read through a pipe is refused.** `npm test | tail -5 && git push` pushes on a red
47
+ suite because a pipeline's status is its last command's; the operators' retrospectives
48
+ record this class in at least 26 repositories. `hygiene.maskedGate` decides, the
49
+ PreToolUse hook refuses with the remedy (`set -o pipefail;` or read the gate's own exit
50
+ code); 12 fixtures in `test/hygiene_test.js`.
51
+ - **One source for operator rules (B-207).** `blocks install|uninstall --agent all` runs one
52
+ operation per installed host (own backup and journal line each), skips absent hosts, and
53
+ its preview is the drift report. `SSHLG:REPOSITORY-HANDOFF` is retired (B-174, decided):
54
+ its rule lives in the operator's own rules block and the hub only removes the old copy.
55
+ - **This repository's entry reads less.** `CLAUDE.md` asks for the retrospective's standing
56
+ section only (about 12 KB), not the 101 KB file; `docs/HANDOFF.md` keeps the current entry
57
+ and an index, older entries moved to `docs/HANDOFF-2026-10-archive.md`.
58
+ - **Board:** B-174 and B-207 closed; B-234 to B-241 record what the audits found outside this
59
+ repository's reach (personal skills, duplicate plugins, per-call hook cost, retro floors in
60
+ other repositories, the empty lesson index, stale clones, the dirty-checkout guard, the
61
+ classifier's status questions).
62
+
1
63
  ## v1.69.0 (2026-10-19): seo-aeo-audit 0.29.0
2
64
 
3
65
  - **Pin: seo-aeo-audit 0.29.0.** `scripts/ai_panel.py` scores captured AI answers per
package/README.md CHANGED
@@ -46,17 +46,17 @@ scripts. No services, no telemetry, no API keys.
46
46
 
47
47
  | Skill | Version | What it does |
48
48
  |---|---|---|
49
- | **[super-ux](https://github.com/ssheleg/super-ux)** | 0.62.0 | Scenario-driven UI development. A versioned design chain in `docs/ux/` — the product vision → personas and jobs → user flows → a screens-and-states map with Figma frames → traced scenarios → evidence-backed audits → fix plans, plus `docs/brand/` for how the product speaks. One `/ux` entry point that reaches every skill, two doc-drift linters and a contract doctor. |
49
+ | **[super-ux](https://github.com/ssheleg/super-ux)** | 0.63.0 | Scenario-driven UI development. A versioned design chain in `docs/ux/` — the product vision → personas and jobs → user flows → a screens-and-states map with Figma frames → traced scenarios → evidence-backed audits → fix plans, plus `docs/brand/` for how the product speaks. One `/ux` entry point that reaches every skill, two doc-drift linters and a contract doctor. |
50
50
  | **[task-pipeline](https://github.com/ssheleg/task-pipeline)** | 1.97.1 | Full-cycle delivery orchestrator. An intake grill interrogates the request into a complete brief, then **ten gated stages** carry it — docs, brainstorm and decompose, spec, plan, build, tests, deploy, post-deploy, wiki, acceptance — refusing to advance until each gate passes. Documentation is a deliverable with its own portable gate, and the retrospective it leaves behind is traceable to the commit that earned each lesson. Ships `project-audit`: a cold-start audit of a whole project — discovery, probes whose unmet needs read `blind` rather than clean, production evidence including one artefact served under two channels, and an HTML report beside a JSON sidecar that says what moved. |
51
- | **[agent-sync](https://github.com/ssheleg/agent-sync)** | 1.22.2 | Several agents, one repository, no collisions. Leases with a TTL so two agents cannot claim the same work, race-free id reservation, a run journal and a generated board — over a pluggable knowledge cloud. The answer to "two sessions just committed over each other". |
52
- | **[make-skill](https://github.com/ssheleg/make-skill)** | 0.32.0 | A skill that builds skills. Create, retrofit, audit and publish agent skills and Claude Code plugins: conformance to the [Agent Skills](https://agentskills.io/specification) open standard, [Anthropic's platform rules](https://platform.claude.com/docs/en/agents-and-tools/agent-skills/overview) (per-surface runtime limits, the Skills API, evals) and the [Claude Code plugin reference](https://code.claude.com/docs/en/plugins-reference), marketplace layout, version sync, validator + CI, every distribution channel, the review checklist for third-party skills, and what a *skill author* must know about MCP and A2A — the protocols themselves live in `agent-stack`. |
53
- | **[sheleg-design](https://github.com/ssheleg/sheleg-design-skill)** | 1.65.0 | The taste layer. Cinematic scroll-driven landing pages — one scroll clock, motion that degrades to calm, WebGL particle formations — plus product-UI style packs each shipping a ready token layer, and the Figma border: tokens as variables, design to code without hand-copied values. |
54
- | **[seo-aeo-audit](https://github.com/ssheleg/seo-aeo-audit)** | 0.29.0 | Evidence-first website audit for search **and** answer engines. Ten tracks from crawl access to AI citation mechanics; every finding carries an observation, every recommendation an evidence tier, and the output is a prioritized change plan plus a link-building brief, not a score. Before a site is built, `/seo-aeo-audit brief` writes the build-time brief that the audit later verifies. |
55
- | **[sheleg-dev](https://github.com/ssheleg/sheleg-dev)** | 0.14.0 | The integration layer a product reaches once it has users: **money in, tracking, sign-in, speed**. Stripe subscription billing reconciled into your own database — webhook idempotency, renewals, seats and proration, refunds, price drift, and the coupon offered at the cancel step, where a `duration=once` discount leaves `subscription.discounts` the moment its invoice finalizes and the offer can be taken every cycle; crypto payments that survive under-payment, duplicate webhooks and rate drift; GA4, Google Ads, Meta and LinkedIn under Consent Mode v2; Google sign-in with the account pre-hijacking guard, and the server-side auth surface behind it; Core Web Vitals work that moves the score rather than the report; and Sentry error tracking wired so it does not forward your own credentials to a third party and so a stack trace names the commit that caused it. |
56
- | **[agent-stack](https://github.com/ssheleg/agent-stack)** | 0.29.0 | Production patterns for agent systems, in four skills. **The orchestrator** — tool-calling loops that survive their own context pressure, pipelines with human checkpoints and resume, provider routing with fallback and health checks, memory architecture — form, function and dynamics, the four-stage retrieval pipeline, and what a layer table alone cannot say, and **the shape of the work decided before the work**: an edge that carries no data is no edge, a plan that declares its dependencies is executed in layers rather than in list order, and a parallel layer gets a checker before the node that consumes it. **The harness** — what the agent is *told*: system prompts at the right altitude, tool descriptions the model can act on, technique choice with a verdict each, workflow versus agent and static versus dynamic, plus a seven-track audit of an agent somebody else built and a scanner for the defects that are mechanically visible. **The evals** — suites that measure whether it actually works: judging trajectories, regression fixtures grown from production, judges calibrated before they are trusted. **The interop layer** — MCP at revision 2026-07-28 and the four features it deprecated, running many servers at once, the registry, A2A 1.0, and the gateway, every reference carrying the spec revision it was read against. Plus the wallet side of reselling LLM access: tiered balances, one markup boundary, two-phase commit against a provider API. |
57
- | **[telegram-dev](https://github.com/ssheleg/telegram-dev)** | 0.3.0 | Telegram, split by the API each surface actually speaks. **`telegram-bots`** — the official HTTP Bot API: `update_id` as the only idempotency key an update carries, the `allowed_updates` default that subscribes you to everything *except* `chat_member` and the two reaction types while answering `ok: true`, the webhook secret header that is the real check, rate limits as a design constraint (1/sec per chat, 20/min per group, ~30/sec bulk), and Telegram Stars with a **ten-second** pre-checkout window and a grant that belongs to `successful_payment`. **`telegram-userbots`** — MTProto and Telethon: the session file as a credential equal to the password, `FloodWaitError` as the API telling you exactly how long to wait, pinning across minor releases that move session and entity-cache behaviour, and the account-ban risk a bot token does not carry — the skill opens with whether a user account is needed at all. **`telegram-miniapps`** — the web layer, whose entire security model is one signed query string: verifying `initData` server-side, the `auth_date` window, the Ed25519 path for a third party, and the SDK package whose name moved while the organisation's own page still fronts the old one. Both the verifier and the delivery invariants ship as runnable fixtures with their mutants. |
58
- | **[xr-dev](https://github.com/ssheleg/xr-dev)** | 0.3.4 | **`quest-lifecycle`** maps the product from platform choice through design, build, testing, Store, monetization, launch and operation. The existing owners cover native OpenXR, Kotlin Spatial/hybrid apps, measured rendering optimization, current-source/tool discovery, commercial and asset readiness, and WebXR/PWA. Every stage names evidence and a next action; Meta engine companions and official Godot XR documentation supply implementation detail. |
59
- | **[web3d-dev](https://github.com/ssheleg/web3d-dev)** | 0.1.4 | Realtime 3D on the web with three.js and React Three Fiber, split by the question. **`web3d-runtime`** — how the scene runs: WebGPURenderer and its WebGL2 fallback, TSL and compute, post-processing, device loss and profiling, led by the failures r186 keeps silent: `render()` before `init()` throws, `positionNode` collapses every instance to a point, `computeAsync` does not wait for the GPU, and about twenty drei components do not run on WebGPU. **`web3d-assets`** — the glTF pipeline from a source to the frame within a written budget: gltf-transform and gltfpack, KTX2 and meshopt, the loader that reads the result, sourcing and the four licence questions; it delegates orders to Asset Foundry when a machine runs it and works by the same principles when not. **`web3d-animation`** — a two-sided animation protocol with a shared clip canon: rig, clips, blending, retargeting, root motion, crowds and vertex-animation textures. Every fact is pinned to three@0.186.1, and the gate checks each code block against that release's exports. |
51
+ | **[agent-sync](https://github.com/ssheleg/agent-sync)** | 1.23.0 | Several agents, one repository, no collisions. Leases with a TTL so two agents cannot claim the same work, race-free id reservation, a run journal and a generated board — over a pluggable knowledge cloud. The answer to "two sessions just committed over each other". |
52
+ | **[make-skill](https://github.com/ssheleg/make-skill)** | 0.33.0 | A skill that builds skills. Create, retrofit, audit and publish agent skills and Claude Code plugins: conformance to the [Agent Skills](https://agentskills.io/specification) open standard, [Anthropic's platform rules](https://platform.claude.com/docs/en/agents-and-tools/agent-skills/overview) (per-surface runtime limits, the Skills API, evals) and the [Claude Code plugin reference](https://code.claude.com/docs/en/plugins-reference), marketplace layout, version sync, validator + CI, every distribution channel, the review checklist for third-party skills, and what a *skill author* must know about MCP and A2A — the protocols themselves live in `agent-stack`. |
53
+ | **[sheleg-design](https://github.com/ssheleg/sheleg-design-skill)** | 1.66.0 | The taste layer. Cinematic scroll-driven landing pages — one scroll clock, motion that degrades to calm, WebGL particle formations — plus product-UI style packs each shipping a ready token layer, and the Figma border: tokens as variables, design to code without hand-copied values. |
54
+ | **[seo-aeo-audit](https://github.com/ssheleg/seo-aeo-audit)** | 0.30.0 | Evidence-first website audit for search **and** answer engines. Ten tracks from crawl access to AI citation mechanics; every finding carries an observation, every recommendation an evidence tier, and the output is a prioritized change plan plus a link-building brief, not a score. Before a site is built, `/seo-aeo-audit brief` writes the build-time brief that the audit later verifies. |
55
+ | **[sheleg-dev](https://github.com/ssheleg/sheleg-dev)** | 0.15.0 | The integration layer a product reaches once it has users: **money in, tracking, sign-in, speed**. Stripe subscription billing reconciled into your own database — webhook idempotency, renewals, seats and proration, refunds, price drift, and the coupon offered at the cancel step, where a `duration=once` discount leaves `subscription.discounts` the moment its invoice finalizes and the offer can be taken every cycle; crypto payments that survive under-payment, duplicate webhooks and rate drift; GA4, Google Ads, Meta and LinkedIn under Consent Mode v2; Google sign-in with the account pre-hijacking guard, and the server-side auth surface behind it; Core Web Vitals work that moves the score rather than the report; and Sentry error tracking wired so it does not forward your own credentials to a third party and so a stack trace names the commit that caused it. |
56
+ | **[agent-stack](https://github.com/ssheleg/agent-stack)** | 0.30.0 | Production patterns for agent systems, in four skills. **The orchestrator** — tool-calling loops that survive their own context pressure, pipelines with human checkpoints and resume, provider routing with fallback and health checks, memory architecture — form, function and dynamics, the four-stage retrieval pipeline, and what a layer table alone cannot say, and **the shape of the work decided before the work**: an edge that carries no data is no edge, a plan that declares its dependencies is executed in layers rather than in list order, and a parallel layer gets a checker before the node that consumes it. **The harness** — what the agent is *told*: system prompts at the right altitude, tool descriptions the model can act on, technique choice with a verdict each, workflow versus agent and static versus dynamic, plus a seven-track audit of an agent somebody else built and a scanner for the defects that are mechanically visible. **The evals** — suites that measure whether it actually works: judging trajectories, regression fixtures grown from production, judges calibrated before they are trusted. **The interop layer** — MCP at revision 2026-07-28 and the four features it deprecated, running many servers at once, the registry, A2A 1.0, and the gateway, every reference carrying the spec revision it was read against. Plus the wallet side of reselling LLM access: tiered balances, one markup boundary, two-phase commit against a provider API. |
57
+ | **[telegram-dev](https://github.com/ssheleg/telegram-dev)** | 0.4.0 | Telegram, split by the API each surface actually speaks. **`telegram-bots`** — the official HTTP Bot API: `update_id` as the only idempotency key an update carries, the `allowed_updates` default that subscribes you to everything *except* `chat_member` and the two reaction types while answering `ok: true`, the webhook secret header that is the real check, rate limits as a design constraint (1/sec per chat, 20/min per group, ~30/sec bulk), and Telegram Stars with a **ten-second** pre-checkout window and a grant that belongs to `successful_payment`. **`telegram-userbots`** — MTProto and Telethon: the session file as a credential equal to the password, `FloodWaitError` as the API telling you exactly how long to wait, pinning across minor releases that move session and entity-cache behaviour, and the account-ban risk a bot token does not carry — the skill opens with whether a user account is needed at all. **`telegram-miniapps`** — the web layer, whose entire security model is one signed query string: verifying `initData` server-side, the `auth_date` window, the Ed25519 path for a third party, and the SDK package whose name moved while the organisation's own page still fronts the old one. Both the verifier and the delivery invariants ship as runnable fixtures with their mutants. |
58
+ | **[xr-dev](https://github.com/ssheleg/xr-dev)** | 0.4.0 | **`quest-lifecycle`** maps the product from platform choice through design, build, testing, Store, monetization, launch and operation. The existing owners cover native OpenXR, Kotlin Spatial/hybrid apps, measured rendering optimization, current-source/tool discovery, commercial and asset readiness, and WebXR/PWA. Every stage names evidence and a next action; Meta engine companions and official Godot XR documentation supply implementation detail. |
59
+ | **[web3d-dev](https://github.com/ssheleg/web3d-dev)** | 0.2.0 | Realtime 3D on the web with three.js and React Three Fiber, split by the question. **`web3d-runtime`** — how the scene runs: WebGPURenderer and its WebGL2 fallback, TSL and compute, post-processing, device loss and profiling, led by the failures r186 keeps silent: `render()` before `init()` throws, `positionNode` collapses every instance to a point, `computeAsync` does not wait for the GPU, and about twenty drei components do not run on WebGPU. **`web3d-assets`** — the glTF pipeline from a source to the frame within a written budget: gltf-transform and gltfpack, KTX2 and meshopt, the loader that reads the result, sourcing and the four licence questions; it delegates orders to Asset Foundry when a machine runs it and works by the same principles when not. **`web3d-animation`** — a two-sided animation protocol with a shared clip canon: rig, clips, blending, retargeting, root motion, crowds and vertex-animation textures. Every fact is pinned to three@0.186.1, and the gate checks each code block against that release's exports. |
60
60
 
61
61
  They compose. `task-pipeline` hands user-facing work to `super-ux` and takes its
62
62
  leases from `agent-sync`; `make-skill` encodes the repo layout the others are
@@ -1122,7 +1122,7 @@ For an explicit local comparison, run
1122
1122
  The source must contain one complete matching outer block. Output contains only
1123
1123
  the selected block's before/after text, so treat that output as local material.
1124
1124
  The command does not apply the diff. Unknown/foreign blocks remain inspectable;
1125
- REPOSITORY-HANDOFF ownership is still a separate operator decision.
1125
+ `SSHLG:REPOSITORY-HANDOFF` is retired (1.70.1): keep its rule in your own block.
1126
1126
 
1127
1127
  ### Write another tool's block (managed-blocks/1)
1128
1128
 
@@ -1133,13 +1133,16 @@ A tool that keeps its own block in a host instruction file — Observatory's
1133
1133
  npx sshlg-skills blocks install OBSERVATORY:AGENT-UPDATES --owner OBSERVATORY --agent claude --from ./block.md
1134
1134
  npx sshlg-skills blocks install OBSERVATORY:AGENT-UPDATES --owner OBSERVATORY --agent claude --from ./block.md --apply
1135
1135
  npx sshlg-skills blocks uninstall OBSERVATORY:AGENT-UPDATES --owner OBSERVATORY --agent claude --apply
1136
+ # one source, every installed host; the preview is the drift report
1137
+ npx sshlg-skills blocks install OPERATOR:GLOBAL-RULES --owner OPERATOR --agent all --from ./rules.md
1136
1138
  ```
1137
1139
 
1138
1140
  Without `--apply` it previews and writes nothing. With it, the file is copied to
1139
1141
  `~/.sshlg-skills/backups/` first, only the named block changes, and each step is
1140
1142
  recorded in `~/.sshlg-skills/managed-blocks/journal.jsonl`. It writes `claude`,
1141
1143
  `codex`, `gemini` and `pi` files, never installs a host, and refuses the hub's own
1142
- blocks (they have their own writers), `SSHLG:REPOSITORY-HANDOFF` (no owner decided)
1144
+ blocks (they have their own writers), any write of the retired
1145
+ `SSHLG:REPOSITORY-HANDOFF` (`blocks uninstall ... --owner SSHLG` removes an old copy)
1143
1146
  and any block whose owner prefix differs from `--owner`. An interrupted write is
1144
1147
  classified with `npx sshlg-skills blocks resolve <operation-id>`. The convention:
1145
1148
  [docs/working-rules/managed-blocks.md](docs/working-rules/managed-blocks.md).
@@ -313,7 +313,7 @@ Usage:
313
313
  npx sshlg-skills visibility marks [--init --show <owner/repo,…> [--replace] [--apply]]
314
314
  # this machine's pack marks
315
315
  # (~/.sshlg-skills/pack-marks.json)
316
- npx sshlg-skills blocks [--agent claude|codex|gemini|pi|cursor] [--json]\n npx sshlg-skills blocks diff OWNER:NAME --from <local-file> --agent <host>\n # read-only markers and selected local diff\n npx sshlg-skills blocks install|uninstall OWNER:NAME --owner OWNER --agent <host> [--from <file>] [--apply]\n npx sshlg-skills blocks resolve <operation-id>\n # managed-blocks/1 writer for a tool's own block; preview unless --apply\n npx sshlg-skills rules [--apply] [--agent codex|gemini|pi|cursor] [--claude-only]\n # opt-in wait rule; preview unless --apply\n npx sshlg-skills rules recovery inspect [--json]\n npx sshlg-skills rules recovery finalize --operation <id> [--apply] [--json]\n # metadata only; never restores host instructions\n npx sshlg-skills domains [--apply | --revert] [--root <dir>] [--no-usage]
316
+ npx sshlg-skills blocks [--agent claude|codex|gemini|pi|cursor] [--json]\n npx sshlg-skills blocks diff OWNER:NAME --from <local-file> --agent <host>\n # read-only markers and selected local diff\n npx sshlg-skills blocks install|uninstall OWNER:NAME --owner OWNER --agent <host>|all [--from <file>] [--apply]\n npx sshlg-skills blocks resolve <operation-id>\n # managed-blocks/1 writer for a tool's own block; preview unless --apply\n npx sshlg-skills rules [--apply] [--agent codex|gemini|pi|cursor] [--claude-only]\n # opt-in wait rule; preview unless --apply\n npx sshlg-skills rules recovery inspect [--json]\n npx sshlg-skills rules recovery finalize --operation <id> [--apply] [--json]\n # metadata only; never restores host instructions\n npx sshlg-skills domains [--apply | --revert] [--root <dir>] [--no-usage]
317
317
  npx sshlg-skills domains --enable|--disable <member> [--project <dir>]
318
318
  # xr-dev, web3d-dev, telegram-dev; hook-bearing members stay global
319
319
  # load only in projects that call for them