@goondocks/myco 0.20.2 → 0.21.1

package/dist/src/agent/prompts/agent.md

@@ -1,6 +1,6 @@
 # Myco Intelligence Agent
 
-You are the Myco intelligence agent. You process captured developer session data to build institutional knowledge. Your job is to read raw session activity, extract meaningful observations, build a knowledge graph, maintain spore lifecycle, and synthesize digest context.
+You are the Myco intelligence agent. You process captured developer session data to build institutional knowledge. Your job is to read raw session activity, extract meaningful observations, maintain spore lifecycle, and synthesize digest context.
 
 You operate on a vault database. The capture layer writes raw data (sessions, prompt batches, activities) without any intelligence. You provide the intelligence — deciding what matters, what connects, and what has changed.
 
@@ -12,17 +12,15 @@ You operate on a vault database. The capture layer writes raw data (sessions, pr
 - **vault_unprocessed** — Get prompt batches not yet processed, ordered by ID. Supports cursor-based pagination via `after_id`.
 - **vault_spores** — List existing spores with filters: `observation_type`, `status` (active/superseded/archived), `agent_id`, `session_id`, or fetch exact spores by `ids` when you need full content for a semantic shortlist.
 - **vault_sessions** — List sessions with optional `status` filter, ordered by most recent.
+- **vault_session_summary_material** — Get compact title/summary material for one session in a single read: current title/summary plus the ordered prompt-batch arc with only user prompts and assistant summaries.
 - **vault_search_fts** — Full-text search across prompt batches and activities using FTS5. Best for keyword matches and finding session content. Params: `query`, `type` (prompt_batch, activity), `limit`.
 - **vault_search_semantic** — Semantic similarity search across embedded vault content (spores, sessions, plans, artifacts). Best for finding conceptually related content and shortlist candidates before reading exact records. Params: `query`, `namespace` (spores, sessions, plans, artifacts — omit to search all), `limit`.
 - **vault_read_digest** — Read current digest extracts. Call with no params for metadata, or with a `tier` number (1500/5000/10000) to read that tier's content.
-- **vault_entities** — List knowledge graph entities with optional `type` and `name` filters. Use to check existing entities before creating new ones.
-- **vault_edges** — List graph edges with optional `source_id`, `target_id`, and `type` filters. Use to check existing relationships before creating edges.
+- **vault_edges** — List lineage edges between sessions, prompt batches, and spores. Use for provenance walks: `FROM_SESSION` (spore→session), `EXTRACTED_FROM` (spore→batch), `HAS_BATCH` (session→batch), `DERIVED_FROM` (wisdom→source spores), `SUPERSEDED_BY` (spore→spore). Filters: `source_id`, `target_id`, `type`.
 
 ### Write Tools
 
 - **vault_create_spore** — Create a new observation. Requires `observation_type` and `content`. Optional: `session_id`, `prompt_batch_id`, `importance` (1-10), `tags`, `context`, `file_path`, `properties` (JSON string, e.g., `'{"consolidated_from": ["id1", "id2"]}'`).
-- **vault_create_entity** — Create or update a knowledge graph node. Requires `type` and `name`. Upserts on (type, name). Optional: `properties` object.
-- **vault_create_edge** — Create a directed relationship between entities. Requires `source_id`, `target_id`, `type`. Optional: `session_id`, `confidence` (0-1), `valid_from`, `properties`.
 - **vault_resolve_spore** — Resolve a spore's lifecycle. Requires `spore_id` and `action` (supersede/archive/merge/split/consolidate). Optional: `new_spore_id`, `reason`, `session_id`.
 - **vault_update_session** — Set a session's `title` and/or `summary`.
 - **vault_set_state** — Store a key-value pair for your cursor and preferences.
@@ -61,37 +59,15 @@ A good spore is specific and captures insight, not activity.
 - Create vague spores like "worked on authentication" — be precise about what was discovered or decided
 - Inflate importance scores — most spores should be 3-6
 
-## Entity Types
+## Lineage Edges (read-only, daemon-created)
 
-Only create an entity when it is referenced by 3+ spores from 2+ different sessions and represents a specific, named thing. Entities are hubs in the knowledge graph — not labels for every concept mentioned.
+The daemon maintains a lineage graph automatically — you do not create edges. Use `vault_edges` to walk provenance:
 
-Good entity names: "DaemonClient", "cursor-based pagination", "Chris"
-Bad entity names: "testing phase", "technical debt", "code quality"
-
-Three types only:
-
-- **component** — A module, class, service, or significant function (e.g., "EventBuffer", "DaemonClient", "SQLite")
-- **concept** — An architectural pattern or domain concept that spans multiple sessions. Must be specific and named, not abstract categories (e.g., "idempotent writes", "cursor-based pagination")
-- **person** — A contributor or team member mentioned in sessions
-
-## Relationship Types
-
-**Semantic edges** (you create these via `vault_create_edge`):
-
-- **RELATES_TO** — General semantic relationship (spore→spore or entity→entity)
-- **SUPERSEDED_BY** — A newer observation replaces an older one (spore→spore)
-- **REFERENCES** — A spore references an entity (spore→entity)
-- **DEPENDS_ON** — Architectural dependency (entity→entity)
-- **AFFECTS** — An observation impacts a component (spore→entity)
-
-**Lineage edges** (created automatically — do NOT create these):
-
-- **FROM_SESSION** — spore → session (auto-created on spore insert)
-- **EXTRACTED_FROM** — spore → batch (auto-created on spore insert)
-- **HAS_BATCH** — session → batch (auto-created on batch insert)
-- **DERIVED_FROM** — wisdom spore → source spore (auto-created on consolidation)
-
-Set `confidence` below 1.0 when the relationship is inferred rather than explicitly stated. Include `session_id` for provenance.
+- **FROM_SESSION** — spore → session (which session produced this observation)
+- **EXTRACTED_FROM** — spore → batch (which prompt batch surfaced it)
+- **HAS_BATCH** — session → batch (structural containment)
+- **DERIVED_FROM** — wisdom spore → source spore (consolidation provenance)
+- **SUPERSEDED_BY** — spore → spore (supersession link)
 
 ## Skill Lifecycle Tools
 
@@ -130,15 +106,13 @@ When running as a single-query task (no phased executor), follow this general se
 2. **Extract** — process batches, create/supersede spores, mark processed, update cursor
 3. **Summarize** — update session titles and summaries for touched sessions
 4. **Consolidate** — search for related spores, create wisdom from 3+ clusters, supersede stale pairs
-5. **Build graph** — create entities (check `vault_entities` first), link with semantic edges (check `vault_edges` first)
-6. **Update digest** — read current tiers via `vault_read_digest`, integrate new material, write updated tiers
-7. **Report** — call `vault_report` with counts and outcomes
+5. **Update digest** — read current tiers via `vault_read_digest`, integrate new material, write updated tiers
+6. **Report** — call `vault_report` with counts and outcomes
 
 For phased tasks, follow only your assigned phase instructions. The executor controls phase sequencing.
 
 **Key rules across all modes:**
 - Supersede rather than duplicate — the vault gets sharper, not bigger
-- Check existing entities/edges before creating to avoid duplicates
 - One observation per spore, specific not vague
 - Report via `vault_report` after each significant action
 - If no work to do, report "skip" with reason and finish

package/dist/src/cli.js

@@ -2,5 +2,5 @@
 import { createRequire as __cr } from 'node:module'; const require = __cr(import.meta.url);
 
 // src/entries/cli.ts
-await import("../cli-GGPWH4UO.js");
+await import("../cli-HSLIG7EX.js");
 //# sourceMappingURL=cli.js.map

package/dist/src/daemon/main.js

@@ -1,6 +1,6 @@
 import { createRequire as __cr } from 'node:module'; const require = __cr(import.meta.url);
 
 // src/entries/daemon.ts
-var { main } = await import("../../main-YTBVRTBI.js");
+var { main } = await import("../../main-4J4QZZTZ.js");
 await main();
 //# sourceMappingURL=main.js.map

package/dist/src/hooks/post-tool-use.js

@@ -1,6 +1,6 @@
 import { createRequire as __cr } from 'node:module'; const require = __cr(import.meta.url);
 
 // src/entries/post-tool-use.ts
-var { main } = await import("../../post-tool-use-POGPTJBA.js");
+var { main } = await import("../../post-tool-use-TZINWWDH.js");
 await main();
 //# sourceMappingURL=post-tool-use.js.map

package/dist/src/hooks/session-end.js

@@ -1,6 +1,6 @@
 import { createRequire as __cr } from 'node:module'; const require = __cr(import.meta.url);
 
 // src/entries/session-end.ts
-var { main } = await import("../../session-end-FT27DWYZ.js");
+var { main } = await import("../../session-end-FS46UARX.js");
 await main();
 //# sourceMappingURL=session-end.js.map

package/dist/src/hooks/session-start.js

@@ -1,6 +1,6 @@
 import { createRequire as __cr } from 'node:module'; const require = __cr(import.meta.url);
 
 // src/entries/session-start.ts
-var { main } = await import("../../session-start-LAFICHII.js");
+var { main } = await import("../../session-start-46KPFV2H.js");
 await main();
 //# sourceMappingURL=session-start.js.map

package/dist/src/hooks/stop.js

@@ -1,6 +1,6 @@
 import { createRequire as __cr } from 'node:module'; const require = __cr(import.meta.url);
 
 // src/entries/stop.ts
-var { main } = await import("../../stop-ZPIKVLH4.js");
+var { main } = await import("../../stop-OUEX6KA4.js");
 await main();
 //# sourceMappingURL=stop.js.map

package/dist/src/hooks/user-prompt-submit.js

@@ -1,6 +1,6 @@
 import { createRequire as __cr } from 'node:module'; const require = __cr(import.meta.url);
 
 // src/entries/user-prompt-submit.ts
-var { main } = await import("../../user-prompt-submit-N36KUPHI.js");
+var { main } = await import("../../user-prompt-submit-APMO6FVU.js");
 await main();
 //# sourceMappingURL=user-prompt-submit.js.map

package/dist/src/mcp/server.js

@@ -1,6 +1,6 @@
 import { createRequire as __cr } from 'node:module'; const require = __cr(import.meta.url);
 
 // src/entries/mcp-server.ts
-var { main } = await import("../../server-AGVYZVP5.js");
+var { main } = await import("../../server-AZJSTQEK.js");
 await main();
 //# sourceMappingURL=server.js.map

package/dist/src/symbionts/manifests/claude-code.yaml

@@ -16,6 +16,10 @@ capture:
     # user-committed plans in the repo.
     - ~/.claude/plans/
     - .claude/plans/
+  planTags:
+    # Ultraplan sends the approved cloud plan back to the terminal by injecting
+    # it into the next user prompt as <ultraplan>...</ultraplan>.
+    - ultraplan
 registration:
   hooksTarget: .claude/settings.json
   mcpTarget: .mcp.json

package/dist/src/symbionts/manifests/opencode.yaml

@@ -21,6 +21,13 @@ registration:
   # opencode.json hosts MCP + permissions under non-standard keys
   mcpTarget: opencode.json
   mcpServersKey: mcp
+  # Opencode prepends /opt/homebrew/bin ahead of ~/.local/bin in the env it
+  # gives MCP children, so `myco-run` resolves to the prod shim regardless
+  # of the user's shell PATH. Bake the runtime.command alias into the MCP
+  # command at install time so the spawn skips PATH resolution entirely.
+  # Every other symbiont resolves `myco-run` via PATH and then uses the
+  # shim's runtime.command walk-up — no install-time substitution needed.
+  substituteRuntimeCommand: true
   settingsTarget: opencode.json
   skillsTarget: .agents/skills
   # No instructionsFile — opencode reads AGENTS.md natively

package/dist/src/symbionts/manifests/pi.yaml

@@ -0,0 +1,22 @@
+name: pi
+displayName: Pi
+binary: pi
+configDir: .pi
+pluginRootEnvVar: PI_PLUGIN_ROOT
+hookFields:
+  sessionId: session_id
+  transcriptPath: transcript_path
+  lastResponse: last_assistant_message
+capture:
+  planDirs:
+    - .pi/plans/
+registration:
+  # Pi uses an extension file, not JSON hook entries
+  hooksTarget: .pi/extensions/myco/index.ts
+  hooksFormat: plugin-file
+  # package.json for the extension's deps
+  pluginPackageTarget: .pi/extensions/myco/package.json
+  # Pi has no native MCP — Myco tools are registered via pi.registerTool()
+  # mcpTarget is omitted; tools are wired in the extension itself.
+  skillsTarget: .agents/skills
+  # No instructionsFile — pi reads AGENTS.md natively via context files

package/dist/src/symbionts/templates/agents-starter.md

@@ -2,4 +2,4 @@
 
 <!-- This starter file was created by Myco. Replace it with your project's rules and conventions. -->
 
-Rules haven't been defined yet. Use the `/rules` skill to generate project rules, or edit this file directly.
+Rules haven't been defined yet. Use the `/myco-rules` skill to add durable project rules, or edit this file directly.

package/dist/src/symbionts/templates/pi/package.json

@@ -0,0 +1,6 @@
+{
+  "dependencies": {
+    "@mariozechner/pi-coding-agent": "*",
+    "@sinclair/typebox": "*"
+  }
+}