opencode-skills-collection 4.0.44 → 4.0.46

package/bundled-skills/delegate-setup/references/schema.md

@@ -0,0 +1,100 @@
+# Fleet schema (`delegate-fleet.v1`)
+
+One concept: **lanes**. A lane names an implementer and optional dials.
+
+## Document
+
+```json
+{
+  "version": "delegate-fleet.v1",
+  "lanes": {
+    "feature": {
+      "implementer": "opencode",
+      "model": "opencode/grok",
+      "variant": "high"
+    },
+    "tests": {
+      "implementer": "grok",
+      "effort": "medium"
+    },
+    "complex": {
+      "implementer": "claude",
+      "effort": "high"
+    }
+  }
+}
+```
+
+- `version` must be `delegate-fleet.v1`.
+- `lanes` is an object keyed by lane name (`[A-Za-z0-9][A-Za-z0-9._-]*`).
+- Every lane **requires** `implementer` (a key from the registry below).
+- Other fields are dials; only dials listed for that implementer are allowed.
+
+## Paths
+
+| Scope | Path |
+| --- | --- |
+| Global | `$XDG_CONFIG_HOME/delegate-skills/config.json` when `XDG_CONFIG_HOME` is set; otherwise `~/.config/delegate-skills/config.json` (`os.homedir()` → `HOME` / `USERPROFILE`) |
+| Project | `<git-root>/.delegate/config.json` |
+
+Project overlays global by **whole-lane replace** (same lane name in project fully replaces the global lane).
+Relays apply a project lane only when its exact config content matches the approval hash written under
+that worktree's Git metadata by an explicitly approved `config.mjs write --scope project`. Cloned or
+later-edited project config fails closed until it is reviewed and written again through `delegate-setup`.
+
+## Implementer keys and dials
+
+| Key | Skill | Binary | Supported dials |
+| --- | --- | --- | --- |
+| `claude` | claude-delegate | `claude` | model, effort, timeout, readOnly |
+| `cline` | cline-delegate | `cline` | provider, model, timeout |
+| `codex` | codex-delegate | `codex` | model, effort, sandbox, timeout, readOnly |
+| `commandcode` | commandcode-delegate | `cmd` | model, effort, timeout, readOnly |
+| `opencode` | opencode-delegate | `opencode` | model, **variant**, timeout, readOnly |
+| `agy` | agy-delegate | `agy` | model, effort, timeout, readOnly |
+| `grok` | grok-delegate | `grok` | model, effort, sandbox, timeout, readOnly |
+| `kimi` | kimi-delegate | `kimi` | model, timeout |
+| `qoder` | qoder-delegate | `qodercli` | model, permissionMode, timeout, readOnly |
+| `vibe` | vibe-delegate | `vibe` | timeout, readOnly |
+| `cursor` | cursor-delegate | `cursor-agent` | model, force, timeout, readOnly |
+| `pi` | pi-delegate | `pi` | provider, model, timeout, readOnly |
+| `omp` | omp-delegate | `omp` | provider, model, **effort** (`--thinking`), timeout, readOnly |
+| `aider` | aider-delegate | `aider` | model, timeout, readOnly |
+| `copilot` | copilot-delegate | `copilot` | model, effort, timeout, readOnly |
+| `warp` | warp-delegate | `oz` | model, timeout |
+| `zcode` | zcode-delegate | `zcode` | permissionMode, timeout, readOnly |
+
+ZCode carries its `--mode` as `permissionMode`, and only `plan` and `yolo` are accepted: ZCode also
+documents `build` and `edit`, but a headless run has no permission client, so those two block every
+write tool and exit 0 having changed nothing. ZCode has no `--model` flag — the model is chosen in
+the CLI's own config file — so `model` is not a dial for `zcode` lanes. ZCode also ships its CLI
+inside the desktop app rather than on PATH, so discovery falls back to the installed app bundle.
+
+OpenCode uses `variant` for reasoning intensity, not `effort`. Do not write `effort` on an `opencode` lane.
+Oh My Pi (`omp`) uses the lane `effort` dial for omp's `--thinking` (`off`, `auto`, `minimal`, `low`, `medium`, `high`, `xhigh`, `max`). Do not write `thinking` as a lane field.
+OpenCode lanes **require** `model` in `provider/model` form, with a non-empty provider before the first
+`/` and at least one non-`/` character after it. Cline accepts `provider` and `model` as separate
+dials and does not impose that shape.
+
+Boolean dials: `readOnly`, `force`. All other dials are non-empty strings. Duration strings for
+`timeout` use `h`/`m`/`s` (e.g. `30m`) and must fit the relay watchdog ceiling (~24.8 days).
+Do not combine `readOnly: true` with a write-capable `sandbox` / `permissionMode` / `force`.
+`model` / `provider` / OpenCode `variant` must match the bound relay’s token rules (e.g. Claude
+rejects spaces; Grok/Pi/Oh My Pi/OpenCode/Codex/Command Code use a shell-safe token set — Windows `shell:true`
+launches, and Oh My Pi's flag-injection defense even without a shell).
+
+## Helpers
+
+```bash
+node <skill-dir>/scripts/discover.mjs
+node <skill-dir>/scripts/config.mjs load [--cwd <dir>]
+node <skill-dir>/scripts/config.mjs validate <file>
+node <skill-dir>/scripts/config.mjs write --scope global|project [--cwd <dir>] <file>
+node <skill-dir>/scripts/lane.mjs resolve --cwd <dir> --lane <name> --implementer <key>
+```
+
+`load` prints the **effective** map (each lane includes a `source` of `global` or `project`) and
+`projectTrusted`, which reports whether the current project content matches its local approval hash.
+`lane.mjs resolve` is what `*-delegate` relays call for `--lane`: it fails loud on a missing
+lane, untrusted project config, or implementer mismatch, and prints relay-native dials
+(e.g. grok `sandbox` → `autonomy`).

package/bundled-skills/delegate-setup/references/setup-dialogue.md

@@ -0,0 +1,101 @@
+# Setup dialogue details
+
+Load this when running a configure / reconfigure session. `SKILL.md` is authoritative and states each
+rule once; this page expands the dialogue itself — the interview, how to ask it, how to read a usage
+scan, and how to label what you propose.
+
+## Grounding the lane map
+
+The grounding menu is `SKILL.md` step 3. When the user picks the interview:
+
+### The four interview questions
+
+Allocation policy only. Users cannot rank model IDs — that is your job, not theirs.
+
+| # | Ask | What it settles |
+| --- | --- | --- |
+| 1 | What kind of work do you delegate most? | The main lane — and its **name**. “migrations”, “bug triage”, “release-prep” beat the canned `feature`/`tests`/`ui`/`fast`/`complex` five |
+| 2 | Which paid subscriptions should I burn, and which should I spare? | Quota economics. Discovery cannot see plans, limits, or what a run costs — only the user knows |
+| 3 | Any CLI you already trust — and for what kind of work? Any that has burned you? | Lived experience outranks your priors about the underlying models. Trust is scoped: an answer qualifies the CLI for work like the work that earned it — follow up on the scope before it qualifies a high-stakes lane. It is never an order to assign work: "trusted + spared" means qualified but normally excluded (the spare rule is in `SKILL.md` step 3) |
+| 4 | Default to fast and cheap, or slow and thorough? | Effort / variant dials, and who gets the `complex` lane |
+
+Stop at four.
+
+### How to ask them
+
+The questions are conversation, not a survey — cramming all four into one cold multiple-choice
+form loses exactly what they exist to collect.
+
+- **One medium per round.** Every question you ask in a turn goes through the same channel: all
+  prose, or all in one form. Never mix the two — submitting a form ends the turn, so any prose
+  question asked beside it is simply lost, and the silence that follows is not an answer to it.
+- **Question 1 is open-ended.** You want the user's words — they become lane names. If your harness
+  forces options, derive them from evidence (repo, usage scan, what the user has said), keep them one
+  genre (kinds of work — never a mix of domains and task types, which overlap), and treat a selection
+  as a draft lane name, not a category.
+- **Questions 2 and 3 are set-valued.** The answer is a mapping across CLIs, so options must span the
+  *discovered* CLIs — never an arbitrary subset — with both directions expressible: burn *and* spare,
+  trust *and* burned-by. Multi-select if the harness has it; otherwise ask in prose.
+- **Question 4 may be a closed choice.** Do not pre-mark an option as recommended — recommendations
+  belong in the proposal, after the answers.
+- **Lead with question 1.** Its answer usually reshapes or removes the others — fewer than four is
+  better.
+- **Silence shrinks the map** — the consent rule is in `SKILL.md` step 3; here is the phrasing for
+  it. Name the axis in one line — “nobody told me which subscriptions to spare, so I set no effort
+  dials; each CLI will use the default you configured” — and add that the answer is welcome anytime.
+
+### Reading a usage scan
+
+`node <skill-dir>/scripts/discover.mjs --usage` adds `usage` to each discovered CLI:
+`{ "sessions": <int>, "lastUsed": <ISO-8601 | null> }`, or `null`.
+
+- What to tell the user before running it: it counts session files and reads their timestamps. It
+  never opens one, so no conversation content is read.
+- `usage: null` also covers a CLI with no local state directory — never report it as zero.
+- Large disparities show where activity occurs — not what role the CLI played in it. 1600 codex
+  sessions against 20 pi sessions cannot tell you whether the user works inside codex (and would
+  orchestrate from it) or sends delegated work to it. Treat a meaningful disparity as a signal to
+  **ask**, never to assign: "codex has the most local sessions, but the scan cannot tell whether
+  you work inside it or delegate work to it — should I protect its quota, burn it as an
+  implementer, or treat it as mixed?" Combine the answer with the burn/spare interview answer when
+  the interview also ran; in a scan-only session the role answer is your only quota evidence —
+  propose conservatively and name the map quota-blind (the shrink rule in `SKILL.md` step 3).
+- Low usage alone is not evidence of task fit. A surplus CLI still needs to be installed,
+  authenticated, and reliable to earn a lane.
+- Small differences are noise. 97 against 61 decides nothing — fall back to the interview or to your
+  opinion, and label it as such.
+- `lastUsed` weighs as much as the count. A big count that stopped months ago is a CLI the user moved
+  off; a recent date on a small count is one they are adopting.
+- Counts are lifetime totals for that machine, not “this month”, and only cover sessions the CLI still
+  keeps on disk.
+
+### Labelling the basis
+
+The Basis values, and the rule that every lane carries one, are in `SKILL.md`. What lives here is the
+split label and the addendum.
+
+Label the parts separately when they differ. A lane whose implementer came from the usage scan but
+whose model you chose is `usage + my opinion (model)`, never a flat `usage data` — session counts
+say where the user works, not which model or effort level to buy for them. That split only exists in
+quick-defaults mode, or when the user asked you for a dial: the evidence modes gate every dial
+(`SKILL.md` rule 7), so there is no opinion-dial left to label.
+
+In the evidence modes, unsolicited opinions about dials travel as an **addendum**, never as a
+pre-filled field (quick-defaults proposals may include opinion-labeled dials — the user hired that
+opinion). Show the table and the JSON first, then, in a separate paragraph after it: “If you want my
+picks for models and effort levels, say the word and I’ll add them.” Then wait. A dial the user
+asked for is theirs; the same dial sitting inside the JSON they are about to approve spent their
+quota on your say-so.
+
+## Auth and models
+
+- `authenticated: null` (unknown) usually means no auth probe is wired for that CLI — currently `agy`
+  and `pi`, which expose no status command. Say that, rather than implying the login failed.
+- Prefer not binding a lane to a CLI discover reports as `authenticated: false`.
+- For claude specifically, `authenticated: false` can be a Keychain artifact when discovery itself
+  ran inside a sandbox: on macOS the live credentials sit in the login Keychain, and a sandbox that
+  blocks Keychain access makes the probe fall back to a possibly stale credentials file. Verify with
+  `claude auth status` outside the sandbox before treating the CLI as unauthenticated.
+- Never invent model ids (rule 6): use `models.values` when `status` is `reported`, or ask the user,
+  or omit `model` when the CLI has a safe default (OpenCode does **not** — require a model for
+  opencode lanes).

package/bundled-skills/docs/integrations/jetski-cortex.md

@@ -1,9 +1,9 @@
 ---
 title: Jetski/Cortex + Gemini Integration Guide
-description: "Use agentic-awesome-skills with Jetski/Cortex without hitting context-window overflow with 2,074+ skills."
+description: "Use agentic-awesome-skills with Jetski/Cortex without hitting context-window overflow with 2,098+ skills."
 ---
 
-# Jetski/Cortex + Gemini: safe integration with 2,074+ skills
+# Jetski/Cortex + Gemini: safe integration with 2,098+ skills
 
 > **Custom-host integration:** This guide documents a low-level, direct-manifest lazy loader for Jetski/Cortex and similar hosts. For Codex or Claude Code, the recommended path is [AAS Core](../users/aas-core.md), which provides neutral, deterministic catalog retrieval and validates exact agent-selected IDs through a bounded, read-only MCP server.
 
@@ -25,7 +25,7 @@ Never do:
 - concatenate all `SKILL.md` content into a single system prompt;
 - re-inject the entire library for **every** request.
 
-With 2,074+ skills, this approach fills the context window before user messages are even added, causing truncation.
+With 2,098+ skills, this approach fills the context window before user messages are even added, causing truncation.
 
 ---
 

package/bundled-skills/docs/integrations/jetski-gemini-loader/README.md

@@ -23,7 +23,7 @@ This example shows one way to integrate **agentic-awesome-skills** with a Jetski
 - How to enforce a **maximum number of skills per turn** via `maxSkillsPerTurn`.
 - How to choose whether to **truncate or error** when too many skills are requested via `overflowBehavior`.
 
-This pattern avoids context overflow when you have 2,074+ skills installed.
+This pattern avoids context overflow when you have 2,098+ skills installed.
 
 Manifest contract references:
 

package/bundled-skills/docs/maintainers/repo-growth-seo.md

@@ -29,7 +29,7 @@ Preferred homepage:
 Preferred social preview:
 
 - lead with `AAS Core` and the profile → stack → plan flow;
-- present `2,074+ Agentic Skills` as supporting catalog evidence, not a second product;
+- present `2,098+ Agentic Skills` as supporting catalog evidence, not a second product;
 - mention Codex and Claude as the current Core agent path, with broader host compatibility as distribution support;
 - avoid dense text and tiny logos that disappear in social cards.
 

package/bundled-skills/docs/maintainers/skills-update-guide.md

@@ -72,7 +72,7 @@ The update process refreshes:
 - Canonical skills index (`skills_index.json`)
 - Compatibility mirror (`data/skills_index.json`)
 - Web app skills data (`apps\web-app\public\skills.json`)
-- All 2,074+ skills from the skills directory
+- All 2,098+ skills from the skills directory
 
 ## When to Update
 

package/bundled-skills/docs/users/aas-core.md

@@ -31,7 +31,7 @@ AAS MCP does not scan the repository and does not decide which skills are best.
 > **Release boundary:** AAS Core landed after release 14.6.0. Use an exact Core-capable release rather than an unreviewed moving tag.
 
 ```bash
-npm exec --yes --ignore-scripts --package=agentic-awesome-skills@16.2.0 -- aas mcp configure \
+npm exec --yes --ignore-scripts --package=agentic-awesome-skills@16.3.0 -- aas mcp configure \
   --host codex \
   --scope user \
   --config /absolute/path/to/codex/config.toml \

package/bundled-skills/docs/users/bundles.md

@@ -1064,4 +1064,4 @@ Found a skill that should be in a bundle? Or want to create a new bundle? [Open
 
 ---
 
-_Last updated: June 2026 | Total Skills: 2,074+ | Total Bundles: 58_
+_Last updated: June 2026 | Total Skills: 2,098+ | Total Bundles: 58_

package/bundled-skills/docs/users/claude-code-skills.md

@@ -17,7 +17,7 @@ Configure AAS Core for Claude Code, describe the task and constraints, let Claud
 - It lets Claude search the verified local catalog without loading the full library into context.
 - It preserves Claude's exact selection without using metadata as an eligibility gate.
 - It keeps MCP discovery read-only and CLI changes approval-gated.
-- It includes 2,074+ skills instead of a narrow single-domain starter pack.
+- It includes 2,098+ skills instead of a narrow single-domain starter pack.
 - It supports the standard `.claude/skills/` path and the Claude Code plugin marketplace flow.
 - It also ships generated bundle plugins so teams can install focused packs like `Essentials` or `Security Developer` from the marketplace metadata.
 - It includes onboarding docs, bundles, and workflows so new users do not need to guess where to begin.

package/bundled-skills/docs/users/gemini-cli-skills.md

@@ -12,7 +12,7 @@ Install into the Gemini skills path, then ask Gemini to apply one skill at a tim
 
 - It installs directly into the expected Gemini skills path.
 - It includes both core software engineering skills and deeper agent/LLM-oriented skills.
-- It helps new users get started with bundles and workflows rather than forcing a cold start from 2,074+ files.
+- It helps new users get started with bundles and workflows rather than forcing a cold start from 2,098+ files.
 - It is useful whether you want a broad internal skill library or a single repo to test many workflows quickly.
 
 ## Install Gemini CLI Skills

package/bundled-skills/docs/users/kiro-integration.md

@@ -18,7 +18,7 @@ Kiro is AWS's agentic AI IDE that combines:
 
 Kiro's agentic capabilities are enhanced by skills that provide:
 
-- **Domain expertise** across 2,074+ specialized areas
+- **Domain expertise** across 2,098+ specialized areas
 - **Best practices** from Anthropic, OpenAI, Google, Microsoft, and AWS
 - **Workflow automation** for common development tasks
 - **AWS-specific patterns** for serverless, infrastructure, and cloud architecture

package/bundled-skills/docs/users/usage.md

@@ -39,7 +39,7 @@ If you came in through a **Claude Code** or **Codex** plugin instead of AAS Core
 
 When you ran `npx agentic-awesome-skills` or cloned the repository, you:
 
-✅ **Downloaded 2,074+ skill files** to your computer (default: `~/.agents/skills/`; or a custom path like `~/.agent/skills/` if you used `--path`)
+✅ **Downloaded 2,098+ skill files** to your computer (default: `~/.agents/skills/`; or a custom path like `~/.agent/skills/` if you used `--path`)
 ✅ **Made them available** to your AI assistant
 ❌ **Did NOT enable them all automatically** (they're just sitting there, waiting)
 
@@ -231,7 +231,7 @@ Let's actually use a skill right now. Follow these steps:
 
 ## Direct-install Step 5: Pick Skills Manually
 
-Don't try to use all 2,074+ skills at once. Here's a sensible approach:
+Don't try to use all 2,098+ skills at once. Here's a sensible approach:
 
 If you want a tool-specific starting point before choosing skills, use:
 
@@ -362,7 +362,7 @@ Usually no, but if your AI doesn't recognize a skill:
 
 ### "Can I load all skills into the model at once?"
 
-No. Even though you have 2,074+ skills installed locally, you should **not** concatenate every `SKILL.md` into a single system prompt or context block.
+No. Even though you have 2,098+ skills installed locally, you should **not** concatenate every `SKILL.md` into a single system prompt or context block.
 
 The intended pattern is:
 

package/bundled-skills/docs/users/visual-guide.md

@@ -40,7 +40,7 @@ agentic-awesome-skills/
 ├── 📄 CONTRIBUTING.md                  ← Contributor workflow
 ├── 📄 CATALOG.md                       ← Full generated catalog
 │
-├── 📁 skills/                          ← 2,074+ skills live here
+├── 📁 skills/                          ← 2,098+ skills live here
 │   │
 │   ├── 📁 brainstorming/
 │   │   └── 📄 SKILL.md                 ← Skill definition
@@ -53,7 +53,7 @@ agentic-awesome-skills/
 │   │   └── 📁 2d-games/
 │   │       └── 📄 SKILL.md             ← Nested skills also supported
 │   │
-│   └── ... (2,074+ total)
+│   └── ... (2,098+ total)
 │
 ├── 📁 apps/
 │   └── 📁 web-app/                     ← Interactive browser
@@ -106,7 +106,7 @@ agentic-awesome-skills/
 
 ```
                     ┌─────────────────────────┐
-                    │  2,074+ SKILLS          │
+                    │  2,098+ SKILLS          │
                     └────────────┬────────────┘
                                  │
         ┌────────────────────────┼────────────────────────┐
@@ -207,7 +207,7 @@ If you want a workspace-style manual install instead, cloning into `.agent/skill
 │   ├── 📁 brainstorming/                 │
 │   ├── 📁 stripe-integration/            │
 │   ├── 📁 react-best-practices/          │
-│   └── ... (2,074+ total)                │
+│   └── ... (2,098+ total)                │
 └─────────────────────────────────────────┘
 ```