@maxkle1nz/m1nd 0.9.0-beta.0

package/skills/m1nd-universal-agent-pack.md

@@ -0,0 +1,81 @@
+# m1nd Universal Agent Pack
+
+Use this pack in any coding-agent host that can read project rules, custom
+instructions, memory, or prompt files.
+
+## Prime Directive
+
+Use `m1nd` as the first investigative layer before raw grep, filesystem
+globbing, or manual file reads when the task depends on repository structure,
+docs, impact, continuity, or risky change preparation.
+
+Skip the first pass only when the user gave exact file/line truth, when compiler
+or runtime output is the only source of truth, or when the task is a trivial
+local file action.
+
+## Startup Trust Loop
+
+1. Call `trust_selftest` if the host exposes it.
+2. If unavailable, call `session_handshake`.
+3. If only `health` is exposed, inspect `tool_surface_contract` and
+   `host_binding_alignment`.
+4. If trust is not full, call or follow `recovery_playbook`.
+5. Only then rely on retrieval surfaces such as `search`, `seek`, or `activate`.
+
+`trust_selftest` and `recovery_playbook` are diagnostic. They do not ingest,
+repair, refresh the host, or mutate the graph.
+
+## Tool Routing
+
+- Exact text -> `search`
+- Path pattern -> `glob`
+- Known purpose, unknown location -> `seek`
+- Topic, subsystem, or neighborhood -> `activate`
+- Unfamiliar repo -> `audit`
+- Stacktrace or runtime error text -> `trace`
+- Risky change -> `impact`, `predict`, `validate_plan`, then usually
+  `surgical_context_v2`
+- Docs/specs -> `ingest` with `adapter="universal"` or `adapter="light"`, then
+  document binding/drift tools
+
+## Recovery Rules
+
+If `seek`, `search`, or `activate` returns `blocked`, zero candidates, or an
+unexpectedly empty graph after ingest, treat it as possible stale binding or
+session split-brain before blaming the repo.
+
+Pass the returned `recovery.arguments` to `recovery_playbook` when present. If
+the response has no payload, call `recovery_playbook` with:
+
+```json
+{"agent_id":"agent","observed_tool":"seek","observed_proof_state":"blocked","observed_candidates":0}
+```
+
+If m1nd is visible but required tools such as `ingest`, `trust_selftest`, or
+`recovery_playbook` are missing, classify the session as
+`degraded_host_tool_surface`. Use m1nd only for orientation and verify final
+truth with local files until the host binding refreshes.
+
+## Change Discipline
+
+Before risky edits:
+
+1. Use `seek` or `activate` to find the connected surface.
+2. Use `impact` for blast radius.
+3. Use `validate_plan` for missing work.
+4. Use `surgical_context_v2` for compact edit context.
+5. Run compiler/tests/runtime checks for execution truth.
+
+`m1nd` complements the compiler, test runner, LSP, debugger, security scanner,
+and local file truth. It does not replace them.
+
+## Continuity
+
+Keep `agent_id` stable within one investigation. Use trails, perspectives, and
+coverage sessions when work spans agents, branches, or sessions.
+
+## L1GHT
+
+Use `adapter="light"` for graph-native semantic markdown. Use
+`adapter="universal"` or `adapter="auto"` for ordinary docs, wiki pages, PDFs,
+and office documents.