@korrlabs/mnemos-pi 2.15.0

package/integrations/skills/mnemos-workflow.md

@@ -0,0 +1,53 @@
+---
+name: mnemos-workflow
+description: Memory workflow lifecycle — track open questions and tasks from open to done without losing them
+---
+
+# Mnemos Workflow
+
+Drive the lifecycle of a memory entry: `open → in-progress → blocked →
+resolved → done` (or `withdrawn`). Use it so open questions and tasks
+survive context resets and multiple agents can pick them up.
+
+## WHEN
+
+- **An open question needs follow-up** — anyone should be able to see its
+  state and who holds it.
+- **Starting work on a tracked item** — move it to `in-progress` first.
+- **Blocked on someone/something** — record WHY in the transition.
+- **Closing work** — `resolved` when answered, `done` when delivered,
+  `withdrawn` when abandoned as no longer relevant.
+
+## STEPS
+
+1. **Check current state** (id from a search result or `mnemos:open-question`
+   entry):
+
+   ```text
+   mnemos_workflow(action="get", memory_id=<id>)
+   ```
+
+2. **Transition with an explicit actor** (your agent slug) and a reason:
+
+   ```text
+   mnemos_workflow(action="set", memory_id=<id>, to="in-progress",
+                   actor=<your-slug>, reason="investigating repro on #123")
+   ```
+
+3. **Review the audit trail** before overriding someone else's state:
+
+   ```text
+   mnemos_workflow(action="history", memory_id=<id>)
+   ```
+
+## RULES
+
+- `blocked → done` is **forbidden** — resolve a blocker first; the state
+  machine enforces this.
+- Stale locks auto-release after 24h; if you force-unlock, give a reason.
+- Same-status transitions are no-ops — don't churn the audit log.
+
+## See also
+
+- Skill `mnemos-write` — creating the `mnemos:open-question` entry first
+- Skill `mnemos-agent-recall` — finding who owns an open item

package/integrations/skills/mnemos-write.md

@@ -0,0 +1,94 @@
+---
+name: mnemos-write
+description: Write a good memory entry — tag contract, content quality, one idea per entry
+---
+
+# Mnemos Write
+
+Persist a single non-session entry (bug-pattern, learning, decision, rule,
+open-question). Use when an agent has discovered something worth keeping
+beyond the current session.
+
+## WHEN
+
+- **When learning something non-obvious** — a gotcha, a hidden constraint, a
+  surprising behaviour that a future agent would benefit from.
+- **When a decision is made** — design, process, or tradeoff chosen, with
+  rationale worth preserving.
+- **When a bug-pattern is identified** — a class of bug that future reviews
+  should catch.
+- **When a rule is established** — a hard constraint or invariant.
+- **When an open question arises** — something that cannot be resolved now
+  but should be revisited.
+
+## STEPS
+
+1. **Choose the tag** (see `mnemos-tag-contract` for the full schema):
+
+   | Tag | When |
+   |-----|------|
+   | `mnemos:bug-pattern` | A class of bug; want future reviews to catch it. |
+   | `mnemos:learning` | An insight to avoid re-learning. |
+   | `mnemos:decision` | A design/process choice + rationale. |
+   | `mnemos:rule` | A hard constraint or invariant. |
+   | `mnemos:open-question` | An unresolved question to revisit. |
+
+2. **Compose the content** — markdown, one idea per entry:
+
+   - For `mnemos:bug-pattern`: describe the bug class, how to spot it, and the
+     fix pattern.
+   - For `mnemos:learning`: state the insight and the context in which it
+     applies.
+   - For `mnemos:decision`: state the decision, the rationale, and the
+     alternatives considered.
+   - For `mnemos:rule`: state the rule and the scope (`applyTo:` tag).
+   - For `mnemos:open-question`: state the question and what is needed to
+     resolve it.
+
+3. **Assemble the tags** (mandatory):
+
+   ```text
+   tags=[
+     "project:<slug>",
+     "agent:<slug>",
+     "mnemos:<subtype>"
+   ]
+   ```
+
+   Add optional tags as needed: `severity:`, `stack:`, `source:`,
+   `applyTo:`, `milestone:`, `domain:`.
+
+4. **Write the entry**:
+
+   ```text
+   mnemos_add(
+     content=<markdown body>,
+     tags=[...],
+     title=<short title>     # optional, auto-generated if omitted
+   )
+   ```
+
+5. **Confirm with a one-line notice**:
+
+   ```text
+   mnemos: wrote <mnemos:subtype> / <title>
+   ```
+
+## DISCIPLINE
+
+- **Do not write trivia.** If you would not want to read this back in 30 days,
+  do not write it. Memory is not a log of every action.
+- **One idea per entry.** Split complex learnings into multiple entries.
+  An entry that covers three topics is unsearchable.
+- **Update by key, don't duplicate.** If re-capturing something, search for it
+  first and update rather than appending a duplicate.
+- **Include the why, not just the what.** A decision without rationale is a
+  rule without justification — future agents cannot tell if it still applies.
+- **Tag contract is mandatory.** Missing or malformed tags cause the write
+  to be rejected in strict mode. See `mnemos-tag-contract`.
+
+## See also
+
+- Skill `mnemos-tag-contract` — full tag schema reference
+- Skill `mnemos-recall` — search before writing (avoid duplicates)
+- Instruction `mnemos-memory-ops.instructions.md`

package/package.json

@@ -0,0 +1,59 @@
+{
+  "name": "@korrlabs/mnemos-pi",
+  "version": "2.15.0",
+  "description": "Mnemos memory & knowledge server — Pi extension. Spawns `mnemos mcp-server` over stdio and exposes every mnemos_* tool as a native Pi tool, plus the mnemos skill pack.",
+  "type": "module",
+  "license": "MIT",
+  "author": "Korrnals",
+  "homepage": "https://github.com/Korrnals/mnemos",
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/Korrnals/mnemos.git"
+  },
+  "bugs": {
+    "url": "https://github.com/Korrnals/mnemos/issues"
+  },
+  "keywords": [
+    "pi-package",
+    "pi",
+    "mcp",
+    "model-context-protocol",
+    "memory",
+    "mnemos",
+    "ai",
+    "coding-agent",
+    "extension"
+  ],
+  "engines": {
+    "node": ">=20"
+  },
+  "pi": {
+    "extensions": [
+      "./integrations/extensions/mnemos-mcp.ts"
+    ],
+    "skills": [
+      "./integrations/skills"
+    ]
+  },
+  "peerDependencies": {
+    "@earendil-works/pi-coding-agent": "*",
+    "typebox": "*"
+  },
+  "peerDependenciesMeta": {
+    "@earendil-works/pi-coding-agent": {
+      "optional": true
+    },
+    "typebox": {
+      "optional": true
+    }
+  },
+  "files": [
+    "integrations/extensions/",
+    "integrations/skills/",
+    "README.md",
+    "LICENSE"
+  ],
+  "publishConfig": {
+    "access": "public"
+  }
+}