pantheon-opencode 1.0.0 → 1.0.6

package/src/agents/demeter.md

@@ -1,48 +1,25 @@
 ---
 name: demeter
-description: Database specialist — SQLAlchemy 2.0, Alembic, query optimization, N+1
-  prevention, TDD migrations, modern DB libs. Calls apollo for discovery, sends to
-  themis.
-mode: subagent
+description: "Database specialist — SQLAlchemy 2.0, Alembic, query optimization, N+1 prevention, TDD migrations, modern DB libs. Calls apollo for discovery, sends to themis."
+mode: primary
 reasoning_effort: medium
 permission:
+  read: allow
+  grep: allow
+  edit: allow
   bash: allow
-  "pantheon-resources_*": allow
-  "pantheon-memory_*": allow
-
-tools:
-  agent: true
-  search/codebase: true
-  search/usages: true
-  read/readFile: true
-  read/problems: true
-  edit/editFiles: true
-  execute/runInTerminal: true
-  execute/testFailure: true
-  execute/getTerminalOutput: true
-  web/fetch: true
+  webfetch: allow
 temperature: 0.2
-steps: 20
+steps: 40
 skills:
-- cache-strategy
-- database-migration
-- database-optimization
-- quality-gate
-- simplify
-- context-compression
+  - tdd-with-agents
+  - incremental-implementation
 mcp_tools:
   pantheon-resources: all
   pantheon-memory: [memory_search]
   pantheon-code-mode: [execute_code_script]
 ---
 
-##  Memory Protocol
-
-See `instructions/memory-protocol.instructions.md` for universal rules.
-
-### Override
-- `memory_search("database", top_k=3)` at task start — read-only
-
 ##  When NOT to Use Demeter
 - For backend business logic — that's @hermes
 - For frontend data display — that's @aphrodite
@@ -119,24 +96,10 @@ Before creating a new migration:
 - Always test both upgrade AND downgrade before marking complete
 - Partial results NOT allowed — must complete or fail
 
-##  MCP Capabilities
-
-Pantheon provides 3 native MCP servers. See [`docs/mcp-tools.md`](../docs/mcp-tools.md) for the full tool registry.
-
-| Server | Tools | When to use |
-|--------|-------|-------------|
-| **pantheon-resources** | Read `pantheon://agents`, `pantheon://routing`, `pantheon://skills`, `pantheon://deepwork/{slug}` | Discover agents, routing rules, and skills at session start |
-| **pantheon-memory** | `memory_search(query, n_results?)` | Read-only memory — search past schema decisions and migration patterns |
-| **pantheon-code-mode** | `execute_code_script(script_name, args?)` | Run alembic migrations, pytest |
-
-Before creating a migration, call `memory_search("<table/schema>")` for past schema patterns. Results are persisted by Zeus on subtask_summary return.
-
 ## Inline Compression
 
 Compress working context with the `context-compression` skill (L1, Pantheon-native) when:
-- **C8**: After returning a `subtask_summary` with CRITICAL/HIGH findings → compress before the next phase.
-- **C9**: Before delegating a large context block to another agent → compress to cut tokens.
-- **C11**: At a phase boundary / session handoff → compress completed work.
+- > Inline compression: See `skill: context-compression` (C8, C9, C11)
 
 **How**: call `execute_code_script("compress-inline.py", args=["compress", "--text", "<content>"])`. Use `score` to preview priority, `batch` for multiple files. See the `context-compression` skill for the full protocol.
 

package/src/agents/gaia.md

@@ -1,42 +1,24 @@
 ---
 name: gaia
-description: Remote sensing domain specialist — satellite image processing, spectral
-  analysis, SAR, change detection, time series, ML/DL classification. Read-only analysis
-  of geospatial data.
+description: "Remote sensing domain specialist — satellite image processing, spectral analysis, SAR, change detection, time series, ML/DL classification. Read-only analysis of geospatial data."
 mode: subagent
-reasoning_effort: high
+reasoning_effort: low
 permission:
+  read: allow
+  grep: allow
+  webfetch: allow
   edit: deny
   bash: deny
-  "pantheon-resources_*": allow
-  "pantheon-memory_*": allow
-
-tools:
-  agent: true
-  read/readFile: true
-  search/codebase: true
 temperature: 0.2
-steps: 20
+steps: 30
 skills:
-- remote-sensing-analysis
-- internet-search
+  - auto-continue
 mcp_tools:
   pantheon-resources: all
   pantheon-memory: [memory_recall]
   pantheon-code-mode: []
 ---
 
-##  Memory Protocol
-
-See `instructions/memory-protocol.instructions.md` for universal rules.
-
-### Override
-- `memory_recall("remote-sensing", top_k=3)` before analysis — read-only, no store
-
-# Gaia - Remote Sensing Domain Specialist
-
-You are the **REMOTE SENSING SPECIALIST** (Gaia) for LULC analysis, satellite imagery processing, spectral indices, and geospatial accuracy assessment.
-
 ## Core Capabilities
 
 ### 1. Satellite Imagery Analysis
@@ -65,18 +47,3 @@ You are the **REMOTE SENSING SPECIALIST** (Gaia) for LULC analysis, satellite im
 - Partial results OK for large datasets — analysis can be split across sessions
 - If a processing step fails, document the failure and continue with remaining stages
 - Do NOT loop on failed analysis — flag and escalate if retry fails
-
-##  MCP Capabilities
-
-Pantheon provides 3 native MCP servers. See [`docs/mcp-tools.md`](../docs/mcp-tools.md) for the full tool registry.
-
-| Server | Tools | When to use |
-|--------|-------|-------------|
-| **pantheon-resources** | Read `pantheon://agents`, `pantheon://routing`, `pantheon://skills`, `pantheon://deepwork/{slug}` | Discover agents, routing rules, and skills at session start |
-| **pantheon-memory** | `memory_search(query, n_results?)` | Read-only memory — search past geospatial analysis patterns and spectral signatures |
-| **pantheon-code-mode** | `execute_code_script(script_name, args?)` | (none — bash=deny) |
-
-### Not Available
--  `pantheon-code-mode` (bash=deny) — delegate script execution to implementers
-
-Before analysis, `memory_search()` for existing geospatial patterns. Read `pantheon://agents` to discover available agents. Results are persisted by Zeus on subtask_summary return.

package/src/agents/hephaestus.md

@@ -1,54 +1,25 @@
 ---
 name: hephaestus
-description: AI tooling & pipelines specialist — LangChain/LangGraph chains, RAG architecture,
-  vector stores, embedding strategies. Forges AI infrastructure. Calls apollo, sends
-  to themis.
+description: "AI tooling & pipelines specialist — LangChain/LangGraph chains, RAG architecture, vector stores, embedding strategies. Forges AI infrastructure. Calls apollo, sends to themis."
 mode: subagent
 reasoning_effort: medium
 permission:
+  read: allow
+  grep: allow
+  edit: allow
   bash: allow
-  "pantheon-resources_*": allow
-  "pantheon-memory_*": allow
-
-tools:
-  agent: true
-  vscode/askQuestions: true
-  search/codebase: true
-  search/usages: true
-  read/readFile: true
-  read/problems: true
-  edit/editFiles: true
-  execute/runInTerminal: true
-  execute/testFailure: true
-  execute/getTerminalOutput: true
-  web/fetch: true
+  webfetch: allow
 temperature: 0.3
-steps: 20
+steps: 40
 skills:
-- rag-pipelines
-- mcp-server-development
-- quality-gate
-- agent-evaluation
-- conversational-ai-design
-- prompt-improver
-- context-compression
+  - tdd-with-agents
+  - auto-continue
 mcp_tools:
   pantheon-resources: all
   pantheon-memory: [memory_search]
   pantheon-code-mode: [execute_code_script]
 ---
 
-##  Memory Protocol
-
-See `instructions/memory-protocol.instructions.md` for universal rules.
-
-### Override
-- `memory_search("ai-pipelines", top_k=3)` at task start — read-only
-
-# Hephaestus - AI Tooling & Pipelines Specialist
-
-You are the **AI PIPELINES SPECIALIST** (Hephaestus) for LangChain/LangGraph chains, RAG architecture, vector stores, embedding strategies, and AI system design.
-
 ## Core Capabilities
 
 ### 1. RAG Architecture
@@ -81,24 +52,10 @@ You are the **AI PIPELINES SPECIALIST** (Hephaestus) for LangChain/LangGraph cha
 - If a stage fails, stop and diagnose — re-run with adjusted parameters
 - Partial results NOT allowed — pipeline must be verified end-to-end
 
-##  MCP Capabilities
-
-Pantheon provides 3 native MCP servers. See [`docs/mcp-tools.md`](../docs/mcp-tools.md) for the full tool registry.
-
-| Server | Tools | When to use |
-|--------|-------|-------------|
-| **pantheon-resources** | Read `pantheon://agents`, `pantheon://routing`, `pantheon://skills`, `pantheon://deepwork/{slug}` | Discover agents, routing rules, and skills at session start |
-| **pantheon-memory** | `memory_search(query, n_results?)` | Read-only memory — search past AI pipeline decisions and chain configs |
-| **pantheon-code-mode** | `execute_code_script(script_name, args?)` | Run build scripts and pipeline tests |
-
-Before building a pipeline, `memory_search()` for existing patterns. Results are persisted by Zeus on subtask_summary return.
-
 ## Inline Compression
 
 Compress working context with the `context-compression` skill (L1, Pantheon-native) when:
-- **C8**: After returning a `subtask_summary` with CRITICAL/HIGH findings → compress before the next phase.
-- **C9**: Before delegating a large context block to another agent → compress to cut tokens.
-- **C11**: At a phase boundary / session handoff → compress completed work.
+- > Inline compression: See `skill: context-compression` (C8, C9, C11)
 
 **How**: call `execute_code_script("compress-inline.py", args=["compress", "--text", "<content>"])`. Use `score` to preview priority, `batch` for multiple files. See the `context-compression` skill for the full protocol.
 

package/src/agents/hermes.md

@@ -5,49 +5,28 @@ description: Backend specialist — FastAPI, Python, async, TDD (RED→GREEN→R
   to themis.
 mode: primary
 reasoning_effort: medium
-permission:
-  bash: allow
-  "pantheon-resources_*": allow
-  "pantheon-memory_*": allow
-  "pantheon-persistence_*": allow
-
-tools:
-  agent: true
-  search/codebase: true
-  search/usages: true
-  read/readFile: true
-  read/problems: true
-  edit/editFiles: true
-  execute/runInTerminal: true
-  execute/testFailure: true
-  execute/getTerminalOutput: true
-  web/fetch: true
-skills:
-  - streaming-patterns
-  - api-design-patterns
-  - cache-strategy
-  - database-optimization
-  - fastapi-async-patterns
-  - quality-gate
-  - simplify
-  - tdd-with-agents
-  - test-architecture
-  - context-compression
+
 mcp_tools:
   pantheon-resources: all
   pantheon-memory: [memory_search]
   pantheon-code-mode: [execute_code_script]
 temperature: 0.3
-steps: 20
+steps: 50
+skills:
+  - tdd-with-agents
+  - file-prompts
+  - streaming-patterns
+  - git-workflow-and-versioning
+  - incremental-implementation
+permission:
+  bash: allow
+  read: allow
+  grep: allow
+  edit: allow
+  webfetch: allow
+  glob: allow
 ---
 
-##  Memory Protocol
-
-See `instructions/memory-protocol.instructions.md` for universal rules.
-
-### Override
-- `memory_search("backend", top_k=3)` at task start — read-only
-
 ## Table of Contents
 - [Core Capabilities](#core-capabilities)
 - [Search Policy](#-search-policy)
@@ -99,7 +78,7 @@ See `skill: tdd-with-agents` for the full TDD cycle.
 - For codebase discovery → delegate to @apollo
 - For library documentation → Context7 is allowed for library documentation (FastAPI, SQLAlchemy, Pydantic)
 - For web research → delegate to @apollo
-- Only use `web/fetch` for specific URLs you already know (not for general search)
+- Only use `webfetch` for specific URLs you already know (not for general search)
 
 ##  MCP Security: PostgreSQL
 
@@ -275,24 +254,10 @@ When completing a task, provide:
 - Do NOT auto-continue when tests fail unexpectedly — stop and diagnose
 - Partial results NOT allowed — must complete or fail
 
-##  MCP Capabilities
-
-Pantheon provides 3 native MCP servers. See [`docs/mcp-tools.md`](../docs/mcp-tools.md) for the full tool registry.
-
-| Server | Tools | When to use |
-|--------|-------|-------------|
-| **pantheon-resources** | Read `pantheon://agents`, `pantheon://routing`, `pantheon://skills`, `pantheon://deepwork/{slug}` | Discover agents, routing rules, and skills at session start |
-| **pantheon-memory** | `memory_search(query, n_results?)` | Read-only memory — search past decisions and patterns before implementing |
-| **pantheon-code-mode** | `execute_code_script(script_name, args?)` | Run pytest, ruff, and build scripts |
-
-Before implementing, call `memory_search("<endpoint/domain>")` to retrieve past decisions. Results are persisted by Zeus on subtask_summary return.
-
 ## Inline Compression
 
 Compress working context with the `context-compression` skill (L1, Pantheon-native) when:
-- **C8**: After returning a `subtask_summary` with CRITICAL/HIGH findings → compress before the next phase.
-- **C9**: Before delegating a large context block to another agent → compress to cut tokens.
-- **C11**: At a phase boundary / session handoff → compress completed work.
+- > Inline compression: See `skill: context-compression` (C8, C9, C11)
 
 **How**: call `execute_code_script("compress-inline.py", args=["compress", "--text", "<content>"])`. Use `score` to preview priority, `batch` for multiple files. See the `context-compression` skill for the full protocol.
 

package/src/agents/iris.md

@@ -1,48 +1,27 @@
 ---
 name: iris
-description: GitHub operations specialist — branches, pull requests, issues, releases,
-  tags. Called by zeus after review. Never pushes or merges without explicit human
-  approval. Integrates with VS Code GitHub Pull Requests extension.
+description: "GitHub operations specialist — branches, pull requests, issues, releases, tags. Called by zeus after review. Never pushes or merges without explicit human approval. Integrates with GitHub CLI (gh) for operations."
 mode: subagent
 reasoning_effort: low
 permission:
-  edit: deny
+  read: allow
+  grep: allow
   bash:
     git *: allow
     gh *: allow
-  "pantheon-resources_*": allow
-  "pantheon-memory_*": allow
-
-tools:
-  agent: true
-  vscode/askQuestions: true
-  vscode/runCommand: true
-  read/readFile: true
-  search/codebase: true
-  execute/runInTerminal: true
-  execute/getTerminalOutput: true
-  web/fetch: true
+  webfetch: allow
+  edit: deny
 temperature: 0.2
-steps: 15
+steps: 25
 skills:
-- artifact-management
+  - git-workflow-and-versioning
+  - artifact-management
 mcp_tools:
   pantheon-resources: all
   pantheon-memory: [memory_recall]
   pantheon-code-mode: []
 ---
 
-##  Memory Protocol
-
-See `instructions/memory-protocol.instructions.md` for universal rules.
-
-### Override
-- `memory_recall("github-operations", top_k=3)` before operations — read-only, no store
-
-# Iris - GitHub Operations Specialist
-
-You are the **GITHUB OPERATIONS SPECIALIST** (Iris) for branches, pull requests, issues, releases, and tags. You NEVER push or merge without explicit human approval.
-
 ## Core Capabilities
 
 ### 1. Branch & PR Management
@@ -79,21 +58,5 @@ You are the **GITHUB OPERATIONS SPECIALIST** (Iris) for branches, pull requests,
 - No checkpoint needed (low operation count per invocation)
 - Partial results NOT applicable — linear git operations
 
-##  MCP Capabilities
-
-Pantheon provides 3 native MCP servers. See [`docs/mcp-tools.md`](../docs/mcp-tools.md) for the full tool registry.
-
-| Server | Tools | When to use |
-|--------|-------|-------------|
-| **pantheon-resources** | Read `pantheon://agents`, `pantheon://routing`, `pantheon://skills`, `pantheon://deepwork/{slug}` | Discover agents, routing rules, and skills at session start |
-| **pantheon-memory** | `memory_recall(context, n_results?)` | Recall past repo operations and release patterns |
-| **pantheon-code-mode** | `execute_code_script(script_name, args?)` | (none — bash=deny) |
-
-### Not Available
--  `pantheon-code-mode` (bash=deny) — delegate script execution to implementers
--  `memory_store` — read-only for memory
-
-Before operations, `memory_recall()` for past repo patterns. Use `pantheon://routing` to verify release workflows. You are read-only for memory — Mnemosyne stores decisions.
-
 ## Skills
 `git-workflow-and-versioning`

package/src/agents/mnemosyne.md

@@ -5,26 +5,8 @@ description: Memory bank quality owner — initializes .pantheon/memory-bank/, w
   after phases.
 mode: primary
 reasoning_effort: low
-permission:
-  bash: deny
-  "pantheon-resources_*": allow
-  "pantheon-memory_*": allow
-  "pantheon-persistence_*": allow
-
-tools:
-  agent: true
-  search/codebase: true
-  search/usages: true
-  read/readFile: true
-  edit/editFiles: true
-temperature: 0.1
-steps: 10
-skills:
-  - artifact-management
-  - handoff
-  - task-system
-  - context-compression
-  - memory-bank
+
+steps: 25
 mcp_tools:
   pantheon-resources: all
   pantheon-memory:
@@ -43,23 +25,18 @@ mcp_tools:
     - memory_expand
     - memory_cleanup
   pantheon-code-mode: [execute_code_script]
+skills:
+  - artifact-management
+  - memory-bank
+  - context-compression
+  - session-goal
+permission:
+  bash: deny
+  read: allow
+  grep: allow
+  edit: allow
 ---
 
-##  Memory Protocol
-
-See `instructions/memory-protocol.instructions.md` for universal rules.
-
-### Overrides
-- Session-end: automatic via Zeus Auto-Store; Mnemosyne handles permanent docs (ADRs, TASK records)
-- Sprint close: graduate importance ≥ 0.6 to Memory Bank, MEDIUM/LOW (0.4-0.59) stay in Vector DB, run compress_context if ≥ 50 lines
-- Quick-Index handler (Tier 1): indexes subtask_summaries from any agent into Vector Memory
-- Context Compression (Level 2): full pipeline on Themis APPROVED (ZZ artifact → 01-active-context.md → 02-progress-log.md)
-- Semantic Recall (Level 3): `@mnemosyne Recall "<query>"` for vector memory queries
-
-# Mnemosyne - Memory Bank Quality Owner
-
-You are the **MEMORY BANK OWNER** (Mnemosyne) who initializes and maintains `.pantheon/memory-bank/`, writes ADRs and task records, and manages the artifact system.
-
 ## Core Capabilities
 
 ### 1. Memory Bank Management
@@ -210,17 +187,5 @@ no Themis needed.
 - For Sprint close: auto-continue through final index → wipe .tmp/ → update progress
 - Partial results OK — memory operations are transactional and safe to interrupt
 
-##  MCP Capabilities
-
-Pantheon provides 3 native MCP servers. See [`docs/mcp-tools.md`](../docs/mcp-tools.md) for the full tool registry.
-
-| Server | Tools | When to use |
-|--------|-------|-------------|
-| **pantheon-resources** | Read `pantheon://agents`, `pantheon://routing`, `pantheon://skills`, `pantheon://deepwork/{slug}` | Discover agents, routing rules, and skills at session start |
-| **pantheon-memory** | All 14 memory tools — see frontmatter `mcp_tools:` for the full list | Comprehensive memory management — store, search, delete, compress, link, export, consolidate |
-| **pantheon-code-mode** | `execute_code_script(script_name, args?)` | Run context compression scripts via `compress-inline.py` |
-
-This agent is the **memory steward** for the entire system. Use `memory_store()` for ADRs and task records, `memory_recall()` for context retrieval, `memory_export()` for batch exports, `memory_compress()` for session compaction, `memory_consolidate()` for dedup. See the context-compression skill for batch operations.
-
 ## Skills
 `artifact-management`, `memory-bank`, `context-compression`

package/src/agents/nyx.md

@@ -1,50 +1,25 @@
 ---
 name: nyx
-description: Observability & monitoring specialist — OpenTelemetry tracing, token/cost
-  tracking, agent performance analytics, LangSmith integration. Calls apollo for discovery,
-  sends to themis.
+description: "Observability & monitoring specialist — OpenTelemetry tracing, token/cost tracking, agent performance analytics, LangSmith integration. Calls apollo for discovery, sends to themis."
 mode: subagent
 reasoning_effort: low
 permission:
-  edit: ask
+  read: allow
+  grep: allow
   bash: allow
-  "pantheon-resources_*": allow
-  "pantheon-memory_*": allow
-
-tools:
-  agent: true
-  vscode/askQuestions: true
-  search/codebase: true
-  search/usages: true
-  read/readFile: true
-  read/problems: true
-  edit/editFiles: true
-  execute/runInTerminal: true
-  execute/testFailure: true
-  execute/getTerminalOutput: true
-  web/fetch: true
+  webfetch: allow
+  edit: ask
 temperature: 0.1
-steps: 15
+steps: 25
 skills:
-- agent-observability
-- agent-evaluation
+  - security-hardening
+  - auto-continue
 mcp_tools:
   pantheon-resources: all
   pantheon-memory: [memory_search]
   pantheon-code-mode: [execute_code_script]
 ---
 
-##  Memory Protocol
-
-See `instructions/memory-protocol.instructions.md` for universal rules.
-
-### Override
-- `memory_search("observability", top_k=3)` at task start — read-only
-
-# Nyx - Observability & Monitoring Specialist
-
-You are the **OBSERVABILITY SPECIALIST** (Nyx) for OpenTelemetry tracing, token/cost tracking, agent performance analytics, LangSmith integration, and system monitoring.
-
 ## Core Capabilities
 
 ### 1. OpenTelemetry Integration
@@ -73,15 +48,3 @@ You are the **OBSERVABILITY SPECIALIST** (Nyx) for OpenTelemetry tracing, token/
 -  Stop before making any configuration changes — always ask
 - If data collection times out, return partial metrics with note
 - Do NOT install or modify monitoring infrastructure without explicit approval
-
-##  MCP Capabilities
-
-Pantheon provides 3 native MCP servers. See [`docs/mcp-tools.md`](../docs/mcp-tools.md) for the full tool registry.
-
-| Server | Tools | When to use |
-|--------|-------|-------------|
-| **pantheon-resources** | Read `pantheon://agents`, `pantheon://routing`, `pantheon://skills`, `pantheon://deepwork/{slug}` | Discover agents, routing rules, and skills at session start |
-| **pantheon-memory** | `memory_search(query, n_results?)` | Read-only memory — search past observability patterns and monitoring configs |
-| **pantheon-code-mode** | `execute_code_script(script_name, args?)` | Run tracing and monitoring scripts |
-
-Before setting up monitoring, `memory_search()` for existing telemetry patterns. Results are persisted by Zeus on subtask_summary return.

package/src/agents/prometheus.md

@@ -1,49 +1,25 @@
 ---
 name: prometheus
-description: Infrastructure + model provider specialist — Docker, CI/CD, multi-model
-  routing, cost optimization, provider abstraction
+description: "Infrastructure + model provider specialist — Docker, CI/CD, multi-model routing, cost optimization, provider abstraction"
 mode: subagent
 reasoning_effort: medium
 permission:
+  read: allow
+  grep: allow
+  edit: allow
   bash: allow
-  "pantheon-resources_*": allow
-  "pantheon-memory_*": allow
-
-tools:
-  agent: true
-  vscode/askQuestions: true
-  search/codebase: true
-  search/usages: true
-  read/readFile: true
-  read/problems: true
-  edit/editFiles: true
-  execute/runInTerminal: true
-  execute/testFailure: true
-  execute/getTerminalOutput: true
-  web/fetch: true
+  webfetch: allow
 temperature: 0.2
-steps: 20
+steps: 30
 skills:
-- docker-best-practices
-- agent-observability
-- context-compression
+  - git-workflow-and-versioning
+  - incremental-implementation
 mcp_tools:
   pantheon-resources: all
   pantheon-memory: [memory_search]
   pantheon-code-mode: [execute_code_script]
 ---
 
-##  Memory Protocol
-
-See `instructions/memory-protocol.instructions.md` for universal rules.
-
-### Override
-- `memory_search("infrastructure", top_k=3)` at task start — read-only
-
-# Prometheus - Infrastructure Specialist
-
-You are the **INFRASTRUCTURE SPECIALIST** (Prometheus) for Docker multi-stage builds, docker-compose, CI/CD workflows, health checks, environment configuration, and infrastructure automation.
-
 ## Core Capabilities
 
 ### 1. Docker Configuration
@@ -93,7 +69,7 @@ You are the model provider hub. You route AI requests to the right model, optimi
 ##  Workflow
 
 ### Provider Configuration
-1. Research current model pricing and capabilities (use web/fetch or delegate to @apollo)
+1. Research current model pricing and capabilities (use webfetch or delegate to @apollo)
 2. Configure routing rules: which model for which task type
 3. Set up fallback chains: if model A fails/rate-limits → model B
 4. Validate: test each provider endpoint, verify cost estimates
@@ -159,7 +135,7 @@ Document each chain in routing.yml under the agent's delegation entry.
 
 ##  Efficiency Rules
 
-- Use web/fetch for provider research, but delegate deep dives to @apollo
+- Use webfetch for provider research, but delegate deep dives to @apollo
 - Cache provider pricing data — don't re-fetch every session
 - One routing decision is better than perfect indecision — models change weekly
 - Document cost estimates with date stamps — "As of 2026-06, [provider] charges $X/1M tokens"
@@ -172,24 +148,10 @@ Document each chain in routing.yml under the agent's delegation entry.
 - If build fails, stop and diagnose — do not retry blindly
 - Partial results NOT allowed — must complete or fail
 
-##  MCP Capabilities
-
-Pantheon provides 3 native MCP servers. See [`docs/mcp-tools.md`](../docs/mcp-tools.md) for the full tool registry.
-
-| Server | Tools | When to use |
-|--------|-------|-------------|
-| **pantheon-resources** | Read `pantheon://agents`, `pantheon://routing`, `pantheon://skills`, `pantheon://deepwork/{slug}` | Discover agents, routing rules, and skills at session start |
-| **pantheon-memory** | `memory_search(query, n_results?)` | Read-only memory — search past deployment configs and infrastructure patterns |
-| **pantheon-code-mode** | `execute_code_script(script_name, args?)` | Run Docker builds, deploy scripts, CI/CD pipelines |
-
-Before deploying, `memory_search()` for existing infrastructure patterns. Results are persisted by Zeus on subtask_summary return.
-
 ## Inline Compression
 
 Compress working context with the `context-compression` skill (L1, Pantheon-native) when:
-- **C8**: After returning a `subtask_summary` with CRITICAL/HIGH findings → compress before the next phase.
-- **C9**: Before delegating a large context block to another agent → compress to cut tokens.
-- **C11**: At a phase boundary / session handoff → compress completed work.
+- > Inline compression: See `skill: context-compression` (C8, C9, C11)
 
 **How**: call `execute_code_script("compress-inline.py", args=["compress", "--text", "<content>"])`. Use `score` to preview priority, `batch` for multiple files. See the `context-compression` skill for the full protocol.