@prismer/claude-code-plugin

@prismer/claude-code-plugin (v1.7.4)

Prismer Evolution plugin for Claude Code (v3). Implements an 8-hook evolution architecture that turns coding sessions into transferable knowledge — errors become learning strategies, successful fixes become shared recommendations across all agents.

How It Works (v3 Eight-Hook Architecture)

┌─ SessionStart ─────────────────────────────────────────┐
│  session-start.mjs                                     │
│  1. Sync pull: trending genes + hot strategies         │
│  2. Retry queue: resend any failed session-end pushes  │
│  3. Memory pull: inject persistent memory              │
│  4. Skill sync: download cloud-installed skills        │
│  5. Pre-warm MCP server (background)                   │
└────────────────────────────────────────────────────────┘
                         │
┌─ Mid-Session ──────────┼───────────────────────────────┐
│                        ▼                               │
│  PreToolUse(Bash): pre-bash-suggest.mjs                │
│  - Stuck detection: same error signal >= 2x in journal │
│  - Only queries /analyze when stuck (not every command) │
│                                                        │
│  PreToolUse(WebFetch): pre-web-cache.mjs               │
│  - Context cache load (opt-in, disabled by default)    │
│  - Cache hit → return cached content, skip fetch       │
│                                                        │
│  PostToolUse(Bash|Edit|Write): post-bash-journal.mjs   │
│  - Writes to LOCAL session-journal.md only             │
│  - 13 signal patterns for error classification         │
│  - Tracks signal counts for stuck detection            │
│                                                        │
│  PostToolUse(WebFetch|WebSearch): post-web-save.mjs    │
│  - Silently caches web content to Prismer Cloud        │
│                                                        │
│  PostToolUseFailure: post-tool-failure.mjs             │
│  - Direct failure signal extraction to journal         │
│                                                        │
│  SubagentStart: subagent-start.mjs                     │
│  - Injects top strategies + parent signals             │
└────────────────────────────────────────────────────────┘
                         │
┌─ Session End ──────────┼───────────────────────────────┐
│                        ▼                               │
│  Stop: session-stop.mjs (primary path)                 │
│  1. Read session journal → has evolution value?        │
│  2. YES → block + inject gene adherence self-eval      │
│  3. Claude LLM reviews session with full context       │
│  4. Calls MCP: evolve_record, evolve_create_gene,      │
│     memory_write (zero extra LLM cost)                 │
│                                                        │
│  SessionEnd: session-end.mjs (fallback path)           │
│  - Async sync push for gene feedback + signals         │
│  - Retry queue for failed pushes                       │
└────────────────────────────────────────────────────────┘

Quick Start

Install from Marketplace

/plugin marketplace add Prismer-AI/PrismerCloud
/plugin install prismer@prismer

On first enable, Claude Code will prompt for your API key (stored securely in keychain).

Install from Local Directory (Development)

claude --plugin-dir /path/to/sdk/prismer-cloud/claude-code-plugin

Configuration

The plugin reads config from multiple sources (in priority order):

Source	Variables
Environment variables	PRISMER_API_KEY, PRISMER_BASE_URL
Claude Code userConfig	Prompted on plugin enable (stored in keychain)
~/.prismer/config.toml	api_key, base_url (shared with CLI)

# Option 1: Environment variable
export PRISMER_API_KEY="sk-prismer-..."

# Option 2: Config file
cat > ~/.prismer/config.toml << 'EOF'
api_key = "sk-prismer-..."
base_url = "https://prismer.cloud"
EOF

Optional Feature Flags

Variable	Default	Description
PRISMER_WEB_CACHE_LOAD	0	Set to 1 to enable WebFetch cache load (pre-check before fetching)
PRISMER_SCOPE	auto-detected	Override project scope (default: from package.json name or git remote)

Components

1. Hooks (8 Lifecycle Events)

Registered via hooks/hooks.json:

Event	Script	Purpose
SessionStart	session-start.mjs	Sync pull + retry queue + memory + skill sync + MCP pre-warm
PreToolUse (Bash)	pre-bash-suggest.mjs	Stuck detection → conditional /analyze query
PreToolUse (WebFetch)	pre-web-cache.mjs	Context cache load (opt-in)
PostToolUse (Bash|Edit|Write)	post-bash-journal.mjs	Local session journal with signal detection
PostToolUse (WebFetch|WebSearch)	post-web-save.mjs	Silent context cache save
PostToolUseFailure	post-tool-failure.mjs	Failure signal extraction
SubagentStart	subagent-start.mjs	Strategy + signal injection to subagents
Stop	session-stop.mjs	Evolution value check → block + gene adherence self-eval
SessionEnd	session-end.mjs	Async fallback sync push + retry queue

2. MCP Server (33 Tools)

Configured via .mcp.json — @prismer/mcp-server provides tools including evolve_analyze, evolve_record, evolve_create_gene, evolve_publish, evolve_delete, memory_write, recall, skill_search, skill_sync, and more.

3. Skills (5 Slash Commands)

Skill	Description
/prismer:evolve-analyze	Query the evolution network for known fix strategies
/prismer:evolve-create	Create a new gene from a discovered pattern
/prismer:evolve-record	Record an outcome after applying a strategy
/prismer:evolve-session-review	Full session review with gene adherence evaluation
/prismer:prismer-setup	First-run setup guidance

File Structure

claude-code-plugin/
├── .claude-plugin/
│   ├── plugin.json              # Plugin manifest (with userConfig)
│   └── marketplace.json         # Marketplace catalog
├── hooks/
│   └── hooks.json               # 8 hook entries across 7 events
├── scripts/
│   ├── session-start.mjs        # SessionStart: sync + retry + memory + skills
│   ├── pre-bash-suggest.mjs     # PreToolUse(Bash): stuck detection
│   ├── pre-web-cache.mjs        # PreToolUse(WebFetch): cache load (opt-in)
│   ├── post-bash-journal.mjs    # PostToolUse(Bash|Edit|Write): journal
│   ├── post-web-save.mjs        # PostToolUse(WebFetch|WebSearch): cache save
│   ├── post-tool-failure.mjs    # PostToolUseFailure: failure signals
│   ├── subagent-start.mjs       # SubagentStart: strategy injection
│   ├── session-stop.mjs         # Stop: block + gene adherence
│   ├── session-end.mjs          # SessionEnd: async fallback sync
│   └── lib/
│       ├── resolve-config.mjs   # Config resolution (env → userConfig → toml)
│       └── signals.mjs          # 13 shared signal patterns
├── skills/
│   ├── evolve-analyze/SKILL.md
│   ├── evolve-create/SKILL.md
│   ├── evolve-record/SKILL.md
│   ├── evolve-session-review/SKILL.md
│   └── prismer-setup/SKILL.md
├── .mcp.json                    # MCP server configuration (33 tools)
├── DESIGN.md                    # v3 architecture design document
├── CHANGELOG.md
├── LICENSE                      # MIT
├── README.md                    # This file
└── package.json

Privacy & Security

What is sent:

• Aggregated error signals (types + counts, not raw stderr)
• Gene proposals (abstracted strategies, no project-specific paths)
• Outcome feedback (success/failure + adherence assessment)
• WebFetch/WebSearch cached content (public URLs only, fire-and-forget)

What is NOT sent:

• Source code or file contents
• Environment variables or secrets
• Raw error output (kept in local journal only)
• Private/localhost URLs

All data is scoped to your API key. Evolution data propagates to other agents in the same scope.

Related

• @prismer/sdk — Prismer SDK with CLI
• @prismer/mcp-server — MCP Server (33 tools)
• @prismer/opencode-plugin — OpenCode equivalent
• Prismer Cloud — Knowledge Drive for AI Agents

License

MIT