agentsmesh 0.26.0 → 0.28.0

package/CHANGELOG.md

@@ -1,5 +1,43 @@
 # Changelog
 
+## 0.28.0
+
+### Minor Changes
+
+- e148490: Deterministic pre-edit lesson recall via a `PreToolUse` hook.
+
+  `agentsmesh init --lessons` now wires the recall hook under **both** `PreToolUse` and `PostToolUse`, and `agentsmesh lessons hook` is event-aware (it echoes the harness `hook_event_name`). On targets whose hooks can inject context before a tool call (e.g. Claude Code), matching lessons are now surfaced **before the first edit/command** — guarding the "first touch" the previous PostToolUse-only design left unguarded — with no extra model turn and no compliance dependence. Session dedup keeps each lesson injected at most once per session, and `PostToolUse` remains as the fallback for harnesses that support only post-call context injection. Existing projects pick this up by re-running `agentsmesh init --lessons` (idempotent) and `agentsmesh generate`; behavior is unchanged when no `hook_event_name` is supplied (defaults to `PostToolUse`).
+
+## 0.27.0
+
+### Minor Changes
+
+- c172153: Declare Amazon Q `additionalRules` as **Native** for project scope. AgentsMesh already emits each non-root rule as a separate `.amazonq/rules/<slug>.md` file (which Amazon Q auto-loads) and imports them back, so the capability was under-declared as `none`. Global scope stays `none` because Amazon Q has no global rules directory on disk (`~/.aws/amazonq/` is used only for MCP and agents).
+- 9492875: Declare Augment Code `permissions` as **Native** for global scope (was `none`). Auggie (Augment's CLI) reads tool permissions from the personal `~/.augment/settings.json` `toolPermissions` array (`[{ toolName, permission: { type: allow | deny | ask-user } }]`), mapped to canonical allow / deny / ask. Generation now emits `toolPermissions` into the global settings.json (alongside `mcpServers`/`hooks`) and import reads it back. Project scope stays `none` — per Augment's docs, tool permissions are personal/global and apply only in the CLI, not the IDE extension. (Advanced `webhook-policy` / `script-policy` / `shellInputRegex` entries have no canonical equivalent and are skipped on import.)
+- c172153: Codex CLI hooks are now declared **Native** (project and global), correcting a `partial` mislabel. Codex CLI reads lifecycle hooks from a real on-disk file — `.codex/hooks.json` (project) and `~/.codex/hooks.json` (global) — which AgentsMesh already generates from `.agentsmesh/hooks.yaml` and imports back, so the round-trip was fully working and only the declared capability level was wrong. The support matrix (README + website) now shows Codex CLI hooks as Native.
+- 4d31ea5: Declare Continue `permissions` as **Native** for global scope (was `none`). Continue reads personal tool permissions from `~/.continue/permissions.yaml` (top-level `allow` / `ask` / `exclude` arrays of tool-matcher patterns); AgentsMesh now generates that file in global mode and imports it back, mapping `exclude` ↔ canonical `deny`. Project scope stays `none` — Continue's own spec marks project-level permissions "not implemented yet".
+
+  Also: the descriptor schema now accepts a target's `globalSupport.scopeExtras` as a valid implementation of a settings-backed capability (mcp/hooks/ignore/permissions) at global scope, since `scopeExtras` is the global-only emit path Continue uses.
+
+- eddf184: Declare GitHub Copilot `hooks` as **Native** for project scope (was `partial`). Copilot CLI auto-loads `.github/hooks/*.json` at startup, and AgentsMesh already generates `.github/hooks/agentsmesh.json` (plus wrapper scripts) and imports it back — the round-trip was fully working and only the declared level was wrong. Lint still warns about unsupported hook events and the POSIX-shell wrapper-script requirement. (Global Copilot hooks at `~/.copilot/hooks/*.json` remain a separate follow-up.)
+- b591409: Declare Cursor `additionalRules` as **Native** (both scopes; was `embedded`). AgentsMesh already emits each non-root rule as a separate native `.cursor/rules/<slug>.mdc` file and imports them back — identical to Cline — so the level was under-declared. No generator/importer changes; the round-trip already worked.
+- dcbe440: Declare Factory Droid `commands` as **Native** (both scopes; was `none`). Factory Droid reads custom slash commands from `.factory/commands/*.md` (project) and `~/.factory/commands/*.md` (global) — Markdown with optional `description` / `allowed-tools` frontmatter, filename as the command slug (per docs.factory.ai). AgentsMesh now generates and imports these native command files instead of projecting commands into `.factory/skills/am-command-*/SKILL.md`; `supportsConversion.commands` is removed. Regenerating replaces the old projected command-skills with native `.factory/commands/` files.
+- b591409: Declare Gemini CLI `permissions` as **Native** for global scope (was `none`). Gemini's policy engine loads User-tier `~/.gemini/policies/*.toml`, and AgentsMesh's permissions generator/importer already produce and read that TOML — the global layout was simply suppressing it. Policies now emit in global mode and round-trip. Project scope stays `partial` (workspace-tier policies are disabled upstream).
+- fd7e4a5: Declare Goose `hooks` as **Native** for both scopes (was `none`). Goose supports lifecycle hooks via the Open Plugin Specification — a `hooks/hooks.json` inside a plugin dir, auto-discovered at `.agents/plugins/agentsmesh/hooks/hooks.json` (project) and `~/.agents/plugins/agentsmesh/hooks/hooks.json` (global). The file uses the same wrapped command-hook shape as Codex CLI / Factory Droid (`{ "hooks": { "<Event>": [{ matcher?, hooks: [{ type, command }] }] } }`), so AgentsMesh now generates and imports it through the shared `wrapped-command-hooks` helper. The previous no-op `lintHooks` warning is removed.
+- c94d3cd: Declare Goose `permissions` as **Native** for global scope (was `none`). Goose reads tool permissions from `~/.config/goose/permission.yaml` — a map keyed by category where AgentsMesh owns the `user` block (`always_allow` / `ask_before` / `never_allow`, mapped to canonical allow / ask / deny). Generation emits it via `scopeExtras` and **merge-preserves** every other category (notably Goose's runtime `smart_approve` cache); import reads the `user` block back. Project scope stays `none` (Goose has no project-level permission file), and the project-scope lint warning now points users at the global file instead of claiming permissions are unsupported.
+- d34681a: Declare Pi Agent `commands` as **Native** (both scopes; was `none`). Pi discovers prompt templates from `.pi/prompts/*.md` (project) and `~/.pi/agent/prompts/*.md` (global) — Markdown with optional `description` frontmatter, filename as the command name, `$ARGUMENTS` substitution (per earendil-works/pi). AgentsMesh now generates and imports these native prompt templates instead of projecting commands into `.pi/skills/am-command-*/SKILL.md`; `supportsConversion.commands` is removed (agents are still projected as skills). Also corrects `metadata.officialUrl` to `https://github.com/earendil-works/pi` (was the non-existent `pi-labs/pi-agent`). Canonical `allowedTools` have no Pi equivalent and are dropped on generate.
+- 0063ecc: Declare Roo Code `agents` as **Native** for project scope (was `partial`) by adding the missing importer. AgentsMesh already generated `.roomodes` (canonical agents → Roo `customModes`); it now imports `.roomodes` back into `.agentsmesh/agents/<slug>.md`, completing the round-trip (`slug` → filename, `name`/`description` → frontmatter, `roleDefinition` → body). Global agents stay `partial` because Roo's global custom modes live in VS Code globalStorage, not an AgentsMesh-ownable file.
+
+### Patch Changes
+
+- ed90714: Make Aider actually load the generated `CONVENTIONS.md`. Aider has no auto-discovery for `CONVENTIONS.md` — it must be referenced from `.aider.conf.yml` via the `read:` key — so the rules AgentsMesh emitted were silently ignored. Generation now also emits a project-level `.aider.conf.yml` with `read: [CONVENTIONS.md]`, merge-preserving any existing user config (other keys kept, `read` list unioned). Scoped to project mode (a home-level config's `read:` path semantics differ and are out of scope); the `.aider.conf.yml` is deterministic wiring and is not imported as canonical content.
+- c172153: Fix Amazon Q agent generation/import using the wrong system-prompt key. AgentsMesh wrote and read `systemPrompt`, but Amazon Q's `agent-v1.json` schema uses `prompt` — so generated `.amazonq/cli-agents/*.json` agents silently lost their system prompt. Generation now emits `prompt`; import reads `prompt` and falls back to the legacy `systemPrompt` key so previously generated agent files still round-trip.
+- c172153: Fix Augment Code rule frontmatter to use the official `type` key. AgentsMesh emitted boolean flags (`always_apply: true`, `agent_requested: true`), but Augment Code rules declare their activation via a single `type` field (`type: always_apply` / `type: agent_requested`). Generation now emits `type: …`; import reads `type` and still accepts the legacy boolean keys for backward compatibility.
+- b591409: Fix `agentsmesh import --from crush --global` silently importing no MCP servers (and no hooks). The Crush config reader was scope-blind and always read `<root>/crush.json`; in global scope it now reads `~/.config/crush/crush.json` (Crush's real global config), so global MCP/hooks round-trip as the `native` matrix already claims.
+- c172153: Fix Cursor hooks being emitted in a format Cursor cannot read. AgentsMesh wrote `.cursor/hooks.json` with Claude-style PascalCase event names (`PreToolUse`) and a nested `{ matcher, hooks: [...] }` structure, but Cursor uses camelCase event names (`preToolUse`, `postToolUse`, `beforeSubmitPrompt`, …) and a flat array of hook objects — so generated hooks silently never fired. Generation and import now use Cursor's real event names and flat shape, the canonical↔Cursor mapping round-trips, and a lint warning is emitted for canonical hook events Cursor has no equivalent for.
+- b2ccc64: Fix `agentsmesh import --from factory-droid` silently dropping native droid agents. Factory Droid's `agents` capability was declared **Native** and generation already emitted `.factory/droids/<name>.md`, but the importer had no `agents` spec — so the round-trip the matrix promised never happened (import produced no `.agentsmesh/agents/*`). The descriptor now imports `.factory/droids/` (and the global `~/.factory/droids/`) back into canonical agents via the `agent` preset, completing the round-trip in both scopes.
+- efd0c6f: Fix Factory Droid `hooks.json` format and complete its round-trip. Factory's `hooks.json` nests events under a top-level `"hooks"` key (`{ "hooks": { "PreToolUse": [{ matcher, hooks: [{ type, command }] }] } }`, per docs.factory.ai) — but generation emitted the bare top-level Claude Code shape Factory does not read, and there was no importer at all, so the declared **Native** hooks never round-tripped. Generation now wraps under `hooks` and `agentsmesh import --from factory-droid` reads it back into `.agentsmesh/hooks.yaml` (both scopes). The wrapped command-hook serialize/import logic — identical to Codex CLI's — is now a shared `wrapped-command-hooks` helper that both targets use, removing the duplication.
+
 ## 0.26.0
 
 ### Minor Changes

package/README.md

@@ -112,7 +112,7 @@ The memory is one git-tracked file, `.agentsmesh/lessons/lessons.json`, and ever
 agentsmesh init --lessons && agentsmesh generate   # wire the recall/capture loop once
 ```
 
-`init --lessons` drops a small always-on rule into `.agentsmesh/rules/_root.md` (so every target gets the habit), seeds the full operating manual as a `lessons` skill where supported, and wires a `PostToolUse` recall hook on hook-capable tools; agents without shell access use the matching MCP tools (`lessons_query` / `lessons_add`). Because the graph is a normal git-tracked file, a lesson one agent learns today helps every teammate's agent tomorrow, and every change is reviewable like any other diff.
+`init --lessons` drops a small always-on rule into `.agentsmesh/rules/_root.md` (so every target gets the habit), seeds the full operating manual as a `lessons` skill where supported, and wires a recall hook (a `PreToolUse` first-touch guard plus a `PostToolUse` fallback) on hook-capable tools; agents without shell access use the matching MCP tools (`lessons_query` / `lessons_add`). Because the graph is a normal git-tracked file, a lesson one agent learns today helps every teammate's agent tomorrow, and every change is reviewable like any other diff.
 
 Full walkthrough: [Teach your AI agents with lessons](https://samplexbro.github.io/agentsmesh/guides/lessons/) · [`agentsmesh lessons` reference](https://samplexbro.github.io/agentsmesh/cli/lessons/).
 
@@ -167,7 +167,7 @@ Every command accepts `--global` (operate on `~/.agentsmesh/`) and `--json` (mac
 AgentsMesh generates native config for every major AI coding assistant — plus plugin targets you can ship as standalone npm packages. Native vs. embedded support per feature is tracked in the [supported-tools matrix](https://samplexbro.github.io/agentsmesh/reference/supported-tools/).
 
 <!-- agentsmesh:tool-list -->
-- **CLI agents:** [Aider](https://aider.chat), [Amp](https://ampcode.com), [Claude Code](https://www.anthropic.com/claude-code), [Codex CLI](https://github.com/openai/codex), [Crush](https://github.com/charmbracelet/crush), [Deep Agents CLI](https://github.com/langchain-ai/deepagents), [Gemini CLI](https://github.com/google-gemini/gemini-cli), [Goose](https://block.github.io/goose), [OpenCode](https://opencode.ai), [Pi Agent](https://github.com/pi-labs/pi-agent), [Qwen Code](https://github.com/QwenLM/qwen-code), [Rovo Dev](https://www.atlassian.com/solutions/devops/rovo-dev), [Warp](https://www.warp.dev).
+- **CLI agents:** [Aider](https://aider.chat), [Amp](https://ampcode.com), [Claude Code](https://www.anthropic.com/claude-code), [Codex CLI](https://github.com/openai/codex), [Crush](https://github.com/charmbracelet/crush), [Deep Agents CLI](https://github.com/langchain-ai/deepagents), [Gemini CLI](https://github.com/google-gemini/gemini-cli), [Goose](https://block.github.io/goose), [OpenCode](https://opencode.ai), [Pi Agent](https://github.com/earendil-works/pi), [Qwen Code](https://github.com/QwenLM/qwen-code), [Rovo Dev](https://www.atlassian.com/solutions/devops/rovo-dev), [Warp](https://www.warp.dev).
 - **IDE integrations:** [Amazon Q Developer](https://aws.amazon.com/q/developer), [Antigravity](https://antigravity.google), [Augment Code](https://www.augmentcode.com), [Cline](https://cline.bot), [Continue](https://continue.dev), [GitHub Copilot](https://github.com/features/copilot), [Cursor](https://cursor.com), [Junie](https://www.jetbrains.com/junie), [Kilo Code](https://kilocode.ai), [Kiro](https://kiro.dev), [Roo Code](https://roocode.com), [Trae](https://www.trae.ai), [Windsurf](https://windsurf.com), [Zed](https://zed.dev).
 - **Cloud agent platforms:** [Factory Droid](https://www.factory.ai), [Jules](https://jules.google), [Replit Agent](https://replit.com).
 <!-- /agentsmesh:tool-list -->
@@ -178,12 +178,12 @@ AgentsMesh generates native config for every major AI coding assistant — plus
 | Feature | Aider | Amazon Q Developer | Amp | Antigravity | Augment Code | Claude Code | Cline | Codex CLI | Continue | GitHub Copilot | Crush | Cursor | Deep Agents CLI | Factory Droid | Gemini CLI | Goose | Jules | Junie | Kilo Code | Kiro | OpenCode | Pi Agent | Qwen Code | Replit Agent | Roo Code | Rovo Dev | Trae | Warp | Windsurf | Zed |
 |---|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|
 | Rules | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native |
-| Additional Rules | Embedded | — | Embedded | Native | Native | Native | Native | Native | Native | Native | Embedded | Embedded | Embedded | Embedded | Embedded | Embedded | Embedded | Native | Native | Native | Native | Embedded | Native | Embedded | Native | Embedded | Native | Embedded | Native | Embedded |
-| Commands | — | — | Native | Partial (workflows) | Native | Native | Native (workflows) | Embedded | Native | Native | — | Native | — | — | Native | — | — | Native | Native | Embedded | Native | — | Native | — | Native | — | Native | — | Native (workflows) | — |
-| Agents | — | Native | — | — | Native | Native | Native | Native | — | Native | — | Native | — | Native | Native | — | — | Native | Native | Native | Native | — | Native | — | Partial | — | — | — | Embedded | — |
+| Additional Rules | Embedded | Native | Embedded | Native | Native | Native | Native | Native | Native | Native | Embedded | Native | Embedded | Embedded | Embedded | Embedded | Embedded | Native | Native | Native | Native | Embedded | Native | Embedded | Native | Embedded | Native | Embedded | Native | Embedded |
+| Commands | — | — | Native | Partial (workflows) | Native | Native | Native (workflows) | Embedded | Native | Native | — | Native | — | Native | Native | — | — | Native | Native | Embedded | Native | Native | Native | — | Native | — | Native | — | Native (workflows) | — |
+| Agents | — | Native | — | — | Native | Native | Native | Native | — | Native | — | Native | — | Native | Native | — | — | Native | Native | Native | Native | — | Native | — | Native | — | — | — | Embedded | — |
 | Skills | Native | — | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | — | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native |
 | MCP Servers | — | Native | Native | — | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | — | — | Native | Native | Native | Native | — | Native | — | Native | Native | Native | Native | Partial | Native |
-| Hooks | — | — | Native | Native | Native | Native | Native | Partial | — | Partial | Native | Native | Native | Native | Partial | — | — | — | — | Native | — | — | Native | — | — | — | — | — | Native | — |
+| Hooks | — | — | Native | Native | Native | Native | Native | Native | — | Native | Native | Native | Native | Native | Partial | Native | — | — | — | Native | — | — | Native | — | — | — | — | — | Native | — |
 | Ignore | Native | — | — | — | Native | Native | Native | — | — | — | Native | Native | — | — | Native (settings-embedded) | Native | — | Native | Native | Native | — | — | Native | — | Native | — | Native | — | Native | — |
 | Permissions | — | — | Native | Partial | — | Native | — | — | — | — | Partial | Partial | — | — | Partial | — | — | — | Native | — | Native | — | Native | — | Partial | — | — | Partial | — | — |
 <!-- /agentsmesh:support-matrix:project -->
@@ -194,14 +194,14 @@ AgentsMesh generates native config for every major AI coding assistant — plus
 | Feature | Aider | Amazon Q Developer | Amp | Antigravity | Augment Code | Claude Code | Cline | Codex CLI | Continue | GitHub Copilot | Crush | Cursor | Deep Agents CLI | Factory Droid | Gemini CLI | Goose | Jules | Junie | Kilo Code | Kiro | OpenCode | Pi Agent | Qwen Code | Replit Agent | Roo Code | Rovo Dev | Trae | Warp | Windsurf | Zed |
 |---|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|:-----------:|
 | Rules | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | — | Native | Native | Native | Native | Native | Native | — | Native | Native | Native | — | Native | — |
-| Additional Rules | Embedded | — | Embedded | Embedded | Native | Native | Native | Embedded | Native | Native | Embedded | Embedded | Embedded | Embedded | Embedded | Embedded | — | Embedded | Native | Native | Native | Embedded | Embedded | — | Native | Embedded | Native | — | Partial | — |
-| Commands | — | — | Native | Partial (workflows) | Native | Native | Native (workflows) | Embedded | Native | Native | — | Native | — | — | Native | — | — | Native | Native | Embedded | Native | — | Native | — | Native | — | Native | — | Native (workflows) | — |
+| Additional Rules | Embedded | — | Embedded | Embedded | Native | Native | Native | Embedded | Native | Native | Embedded | Native | Embedded | Embedded | Embedded | Embedded | — | Embedded | Native | Native | Native | Embedded | Embedded | — | Native | Embedded | Native | — | Partial | — |
+| Commands | — | — | Native | Partial (workflows) | Native | Native | Native (workflows) | Embedded | Native | Native | — | Native | — | Native | Native | — | — | Native | Native | Embedded | Native | Native | Native | — | Native | — | Native | — | Native (workflows) | — |
 | Agents | — | Native | — | — | Native | Native | Native | Native | — | Native | — | Native | — | Native | Native | — | — | Native | Native | Native | Native | — | Native | — | Partial | — | — | — | Embedded | — |
 | Skills | Native | — | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | Native | — | Native | Native | Native | Native | Native | Native | — | Native | Native | Native | Native | Native | — |
 | MCP Servers | — | Native | Native | Native | Native | Native | Native | Native | Native | — | Native | Native | Native | Native | Native | Native | — | Native | Native | Native | Native | — | Native | — | Native | Native | Native | — | Native | Native |
-| Hooks | — | — | Native | Native | — | Native | Native | Partial | — | — | Native | Native | Native | Native | Partial | — | — | — | — | — | — | — | Native | — | — | Native | — | — | Native | — |
+| Hooks | — | — | Native | Native | — | Native | Native | Native | — | — | Native | Native | Native | Native | Partial | Native | — | — | — | — | — | — | Native | — | — | Native | — | — | Native | — |
 | Ignore | Native | — | — | — | — | Native | Native | — | — | — | — | Native | — | — | — | Native | — | — | Native | Native | — | — | — | — | Native | — | — | — | Native | — |
-| Permissions | — | — | Native | Partial | — | Native | — | — | — | — | — | — | — | — | — | — | — | Native | Native | — | Native | — | Native | — | Partial | Native | — | Partial | — | — |
+| Permissions | — | — | Native | Partial | Native | Native | — | — | Native | — | — | — | — | — | Native | Native | — | Native | Native | — | Native | — | Native | — | Partial | Native | — | Partial | — | — |
 <!-- /agentsmesh:support-matrix:global -->
 
 See the [full feature matrix](https://samplexbro.github.io/agentsmesh/reference/supported-tools/) for native vs. embedded details and per-tool global paths.