@ainova-systems/intelligence 0.19.0 → 0.21.0

package/packages/sync/rules/intelligence-authoring.md

@@ -48,13 +48,13 @@ The mistakes that actually happen, in order of frequency:
 
 ## Rules
 
-**Scope with `paths:`.** A rule without `paths:` is loaded into *every* session and inlined into `AGENTS.md` — the most expensive context real estate there is. Earn it. A rule that only matters when someone touches the frontend belongs scoped to the frontend.
+**Scope with `paths:`.** A rule without `paths:` is loaded into *every* session and inlined into `AGENTS.md` — the most expensive context real estate there is. Earn it. A rule that only matters when someone touches the frontend belongs scoped to the frontend. Guidance about one tool belongs to that tool: `targets: [claude]` keeps it out of every other tool's context. A rule whose targets leave `agents` out also leaves `AGENTS.md`, and sync names any tool that can then no longer receive it.
 
 **Lead with the positive default, then the invariant.** The model acts on the first instruction it reads, and it follows whatever is named — negation ("never do X") draws attention to X. Reserve absolute language (NEVER, MUST) for true must-nots: safety, security, output format. A judgement call written as a NEVER only teaches the model that NEVERs are negotiable.
 
 **Derive from the code, not from prose that already exists.** Every REQUIRED and every invariant must be backed by something observed in the repository. Documentation is a claim, not evidence — read the file.
 
-**Describe the repository, not a machine.** The OS, the shell, the editor, a local dev stack, personal tooling — these are environment facts. They belong in a personal, gitignored `CLAUDE.md`. A rule is committed and read by everyone, so a shell-specific command or an absolute local path in it is simply wrong for whoever is on another platform.
+**Describe the repository, not a machine.** The OS, the shell, the editor, a local dev stack, personal tooling — these are environment facts. They belong in a local rule, `<content-dir>/rules/<name>.local.md`: Git ignores it, and sync renders it only into outputs Git ignores. A rule is committed and read by everyone, so a shell-specific command or an absolute local path in it is simply wrong for whoever is on another platform.
 
 **Explain the why — but only a why you can back.** A reason the reader can check is what makes an instruction survive a judgement call. An invented reason is worse than none: it sounds like evidence.
 

package/packages/sync/skills/intelligence-learn-from-repository/SKILL.md

@@ -59,9 +59,10 @@ mechanics; this skill supplies repository judgement.
    Treat backed-up legacy root instructions such as `.cursorrules` and an
    instruction-bearing `CLAUDE.md` as migration sources, not permanent parallel
    entry points. Move still-valid shared guidance into project-owned rules; do
-   not restore the original monolith. Only when approved machine-local guidance
-   remains, create a small new gitignored `CLAUDE.md` from that subset after
-   migration. `AGENTS.md` remains the shared root instruction entry point.
+   not restore the original monolith. Approved machine-local guidance becomes a
+   local rule, `<content-dir>/rules/<name>.local.md`, which Git ignores and sync
+   keeps out of committed output. `AGENTS.md` remains the shared root
+   instruction entry point.
 
    Review `.gitignore` and every existing `.vscodeignore`, `.npmignore`, and
    `.dockerignore` against the CLI-managed policy. Detect tracked files which
@@ -95,8 +96,9 @@ explains itself well.
     relevant generated `AGENTS.md`, Cursor rules, Claude rules, and any
     packaging/build file list affected by ignore policy. Verify quarantined
     legacy root entry points remain absent. If the user approved a genuinely
-    machine-local exception, create a concise new gitignored root file from only
-    that content; never copy the backed-up monolith back. Completion requires
+    machine-local exception, write only that content to a local rule
+    (`<content-dir>/rules/<name>.local.md`); never copy the backed-up monolith
+    back. Completion requires
     `IS_STATUS=ok`, a clean final check, and no `onboarding is pending` header
     after accepted migration.
 11. Report what was created, updated, removed, or deliberately kept. Remind the

package/packages/sync/skills/intelligence-review-context/references/audit-checks.md

@@ -9,7 +9,7 @@ Apply these checks to the source groups resolved by the review skill.
 | **Over the cap** | `SKILL.md` over 1000 lines, rule over 500, agent over 200 | `SPLIT` — two artifacts, or move detail into `references/<topic>.md` |
 | **Shared instruction budget** | The measured shared agents output, or `agents-md` in sync's `CONTEXT:` summary, is over 32 KiB (32,768 bytes) | `COMPACT` — this is Intelligence's recommended maximum, not an adapter rejection threshold; use the compaction procedure in `intelligence-review-context` to reduce the owning sources without teaching terse output |
 | **Rule links to a rule** | A markdown link from one rule to another (`R1`) | `UNLINK` — name the rule instead: always-on rules are inlined into `AGENTS.md` and the scoped channels carry only scoped rules, so the link is dead in at least one output |
-| **Machine facts in a rule** | OS, shell, editor or a local absolute path (`R2`) | `MOVE` — these belong in a personal, gitignored `CLAUDE.md`; a rule is committed and read by everyone, including whoever is on another platform |
+| **Machine facts in a rule** | OS, shell, editor or a local absolute path (`R2`) | `MOVE` — these belong in a local rule, `<content-dir>/rules/<name>.local.md`, which Git ignores and sync keeps out of committed output; a rule is committed and read by everyone, including whoever is on another platform |
 | **Literal path or command in a skill** | A path *outside the skill's own folder* baked into a procedure (`R3`) | `PARAMETERIZE` — a skill is *executed*, so a literal path breaks the moment the layout moves; resolve it from a rule or from `<manifest>`. **Exempt:** the skill's own bundle (`references/`, `scripts/`, `assets/` — content is co-located with its skill by default); rules and agents (describing the repository is their job); an example inside an output-format block; the resolution step itself |
 | **Skill with no verification** | Nothing at the end proves the procedure worked (`R4`) | `FLAG` — a procedure that proves nothing is a note, or just the work: give it a verification, or delete it |
 | **Pressure marker** | A heading or section named CRITICAL / MANDATORY / HARD RULE / READ FIRST, or a density of `MUST` / `NEVER` / `ALWAYS` with no reason beside it (`R5`) | `REWRITE` - plain heading, one reason per constraint; when several instructions are each marked critical the marker stops carrying information, so keep emphasis for the one instruction demonstrably under-weighted without it |

package/packages/sync/skills/intelligence-update-context/references/rules.md

@@ -13,6 +13,11 @@
    Architecture, Build and test, Examples, and Patterns to recognize and replace.
    Keep reasons that guide judgment and reference real examples. An accepted
    user preference is identified as a preference, not a claim about existing code.
+5. Guidance that belongs to one machine or one person — the OS, the shell, a
+   local service, a personal habit — goes in a local rule,
+   `<content-dir>/rules/<name>.local.md`. Git ignores it, and sync renders it
+   only into outputs Git ignores, so it never reaches `AGENTS.md` or a teammate.
+   A local rule cannot share a name with a committed one.
 
 For a scoped rule: