Findings & checks · 4 flagged
HIGHLong base64-encoded blob hidden in the skill documentationSS-SKILL-INJECT-B64-PAYLOAD-01 · Prompt injection · README.md
A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.
- Decode the base64 string and confirm it is not an instruction directed at the agent.
- Move any legitimate binary or signature data into a dedicated file (
*.sig,SIGNATURES) outside the documentation.
Trace & refs
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · README.md×3
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Show all 3 locations
- Delete blanket "don't ask / no need to confirm" directives from the skill.
- If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Trace & refs
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
<p align="center"> <a href="https://github.com/Vvkmnn/claude-emporium"><img src="logo/claude-roman.svg" width="300" height="300" alt="roma victrix"></a> </p>
<div align="center"> <h1>claude emporium</h1> <p><em>a (roman) plugin marketplace for claude code</em></p> </div>
<!-- social --> <p align="center"> <a href="https://github.com/Vvkmnn"><img src="https://img.shields.io/github/followers/Vvkmnn?label=%40vvkmnn&style=social" alt="@vvkmnn"></a> <a href="https://github.com/Vvkmnn/claude-emporium"><img src="https://img.shields.io/github/stars/Vvkmnn/claude-emporium?style=social&label=stars" alt="GitHub stars"></a> </p>
<!-- tech --> <p align="center"> <a href="https://opensource.org/licenses/MIT"><img src="https://img.shields.io/badge/License-MIT-yellow.svg" alt="License: MIT"></a> <a href="https://claude.ai/code"><img src="https://img.shields.io/badge/Claude_Code-D97757?logo=claude&logoColor=fff" alt="Claude Code"></a> <a href="https://modelcontextprotocol.io/"><img src="https://img.shields.io/badge/MCP-6_servers-blue" alt="MCP"></a> <a href="https://www.typescriptlang.org/"><img src="https://img.shields.io/badge/TypeScript-007ACC?logo=typescript&logoColor=white" alt="TypeScript"></a> <a href="https://nodejs.org/"><img src="https://img.shields.io/badge/node-%3E%3D20-brightgreen" alt="Node.js"></a> </p>
<!-- npm versions --> <p align="center"> <a href="https://www.npmjs.com/package/claude-praetorian-mcp"><img src="https://img.shields.io/npm/v/claude-praetorian-mcp.svg?label=praetorian" alt="praetorian"></a> <a href="https://www.npmjs.com/package/claude-historian-mcp"><img src="https://img.shields.io/npm/v/claude-historian-mcp.svg?label=historian" alt="historian"></a> <a href="https://www.npmjs.com/package/claude-oracle-mcp"><img src="https://img.shields.io/npm/v/claude-oracle-mcp.svg?label=oracle" alt="oracle"></a> <a href="https://www.npmjs.com/package/claude-gladiator-mcp"><img src="https://img.shields.io/npm/v/claude-gladiator-mcp.svg?label=gladiator" alt="gladiator"></a> <a href="https://www.npmjs.com/package/claude-vigil-mcp"><img src="https://img.shields.io/npm/v/claude-vigil-mcp.svg?label=vigil" alt="vigil"></a> <a href="https://www.npmjs.com/package/claude-orator-mcp"><img src="https://img.shields.io/npm/v/claude-orator-mcp.svg?label=orator" alt="orator"></a> </p>
<p align="center"> <a href="#install">install</a> <a href="#plugins">plugins</a> <a href="#synergy">synergy</a> <a href="#architecture">architecture</a> <a href="#contributing">contributing</a> </p>
<p align="center"> <sup><em>roma victrix</em></sup> </p>
Six plugins that wrap standalone MCP servers with automation hooks, commands, and skills. No code duplication: plugins tell claude _when_ to act, MCPs handle _how_.
╔═╤═══════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════╤═╗
║ │ │ ║
║ │ PLUGINS │ ║
║ │ │ ║
║ │ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ │ ║
║ │ │ PRAETORIAN │ │ HISTORIAN │ │ ORACLE │ │ GLADIATOR │ │ VIGIL │ │ ORATOR │ │ ║
║ │ │ context guard │ │ session memory │ │ tool discovery │ │ learn & adapt │ │ file recovery │ │ prompt rhetoric │ │ ║
║ │ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ │ ║
║ │ │ hooks │ │ hooks │ │ hooks │ │ hooks │ │ hooks │ │ hooks │ │ ║
║ │ │ · pre-plan │ │ · pre-websearch │ │ · pre-plan │ │ · post-error │ │ · pre-bash │ │ · pre-task │ │ ║
║ │ │ · pre-compact │ │ · pre-plan │ │ · post-error │ │ · stop │ │ │ │ │ │ ║
║ │ │ · post-research │ │ · pre-task │ │ │ │ │ │ commands │ │ commands │ │ ║
║ │ │ · subagent-stop │ │ · post-error │ │ commands │ │ commands │ │ · /save-vigil │ │ · /reprompt- │ │ ║
║ │ │ │ │ │ │ · /search-oracle │ │ · /review- │ │ · /restore-vigil │ │ orator │ │ ║
║ │ │ commands │ │ commands │ │ │ │ gladiator │ │ │ │ │ │ ║
║ │ │ · /compact- │ │ · /search- │ │ │ │ │ │ │ │ │ │ ║
║ │ │ praetorian │ │ historian │ │ │ │ │ │ │ │ │ │ ║
║ │ │ · /restore- │ │ │ │ │ │ │ │ │ │ │ │ ║
║ │ │ praetorian │ │ │ │ │ │ │ │ │ │ │ │ ║
║ │ └────────────────────┘ └────────────────────┘ └────────────────────┘ └────────────────────┘ └────────────────────┘ └────────────────────┘ │ ║
║ │ │ │ │ │ │ │ │ ║
║ │ ▼ ▼ ▼ ▼ ▼ ▼ │ ║
║ │ │ ║
║ │ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ │ ║
║ │ │ praetorian-mcp │ │ historian-mcp │ │ oracle-mcp │ │ gladiator-mcp │ │ vigil-mcp │ │ orator-mcp │ │ ║
║ │ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ ├────────────────────┤ │ ║
║ │ │ save_context │ │ search_convos │ │ search │ │ observe │ │ vigil_save │ │ orator_optimize │ │ ║
║ │ │ · snapshot before │ │ · full-text across │ │ · query 17 sources │ │ · record patterns │ │ · named checkpoint │ │ · score 7 dims │ │ ║
║ │ │ compaction │ │ all sessions │ │ in parallel │ │ │ │ │ │ · apply 8 techs │ │ ║
║ │ │ │ │ │ │ │ │ reflect │ │ vigil_list │ │ · restructure │ │ ║
║ │ │ restore_context │ │ get_error_solns │ │ browse │ │ · cluster and │ │ · show checkpoints │ │ │ │ ║
║ │ │ · load previous │ │ · how errors were │ │ · by category or │ │ recommend │ │ │ │ ── ── ── ── ── │ │ ║
║ │ │ session state │ │ resolved │ │ popularity │ │ │ │ vigil_diff │ │ dimensions: │ │ ║
║ │ │ │ │ │ │ │ │ ── ── ── ── ── │ │ · preview changes │ │ clarity │ │ ║
║ │ │ search_compactns │ │ find_similar │ │ sources │ │ storage: │ │ │ │ specificity │ │ ║
║ │ │ · find past saves │ │ · related past │ │ · list registries │ │ .claude/ │ │ vigil_restore │ │ structure │ │ ║
║ │ │ │ │ questions │ │ and status │ │ gladiator/ │ │ · restore files │ │ context │ │ ║
║ │ │ list_compactions │ │ │ │ │ │ │ │ │ │ examples │ │ ║
║ │ │ · browse recent │ │ find_file_context │ │ ── ── ── ── ── │ │ │ │ vigil_delete │ │ constraints │ │ ║
║ │ │ snapshots │ │ · track changes │ │ smithery · glama │ │ │ │ · remove checkpoint│ │ tone │ │ ║
║ │ │ │ │ │ │ npm · github │ │ │ │ │ │ │ │ ║
║ │ │ ── ── ── ── ── │ │ find_tool_pattns │ │ awesome-mcp │ │ │ │ ── ── ── ── ── │ │ ── ── ── ── ── │ │ ║
║ │ │ storage: │ │ · agent workflows │ │ mcp-registry │ │ │ │ storage: │ │ in-memory │ │ ║
║ │ │ .claude/ │ │ │ │ + 11 more │ │ │ │ .claude/ │ │ zero storage │ │ ║
║ │ │ praetorian/ │ │ search_plans │ │ │ │ │ │ vigil/ │ │ │ │ ║
║ │ │ │ │ · past plans │ │ in-memory cache │ │ │ │ │ │ │ │ ║
║ │ │ │ │ │ │ zero storage │ │ │ │ │ │ │ │ ║
║ │ │ │ │ list_recent │ │ │ │ │ │ │ │ │ │ ║
║ │ │ │ │ · recent sessions │ │ │ │ │ │ │ │ │ │ ║
║ │ └────────────────────┘ └────────────────────┘ └────────────────────┘ └────────────────────┘ └────────────────────┘ └────────────────────┘ │ ║
║ │ │ ║
║ │ MCP SERVERS │ ║
╚═╧═══════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════╧═╝install
Each plugin self-configures on install. MCP servers, hooks, commands, skills, and sibling detection all register automatically. All hooks default to enabled. See configuration to disable specific hooks or tune behavior.
Requirements:
Marketplace:
# Add this marketplace
/plugin marketplace add Vvkmnn/claude-emporiumPlugins:
# Each registers its MCP server, skills, hooks, settings, and detects siblings automatically
/plugin install claude-praetorian@claude-emporium
/plugin install claude-historian@claude-emporium
/plugin install claude-oracle@claude-emporium
/plugin install claude-gladiator@claude-emporium
/plugin install claude-vigil@claude-emporium
/plugin install claude-orator@claude-emporiumMCP servers:
# NOT REQUIRED: Each plugin will install their MCP server by themself
# However, you can choose to install just the MCP servers if you like
# Standalone has no hooks or synergy, but each ships with an optionally installable skill
# claude mcp add praetorian -- npx claude-praetorian-mcp # https://github.com/Vvkmnn/claude-praetorian-mcp
# claude mcp add historian -- npx claude-historian-mcp # https://github.com/Vvkmnn/claude-historian-mcp
# claude mcp add oracle -- npx claude-oracle-mcp # https://github.com/Vvkmnn/claude-oracle-mcp
# claude mcp add gladiator -- npx claude-gladiator-mcp # https://github.com/Vvkmnn/claude-gladiator-mcp
# claude mcp add vigil -- npx claude-vigil-mcp # https://github.com/Vvkmnn/claude-vigil-mcp
# claude mcp add orator -- npx claude-orator-mcp # https://github.com/Vvkmnn/claude-orator-mcpplugins
<h3>praetorian</h3>
<a href="https://github.com/Vvkmnn/claude-praetorian-mcp"><img src="logo/claude-praetorian.svg" width="200" height="200" align="left" alt="praetorian"></a>
Context Guard. Saves and restores valuable context before compaction, after research, and when subagents complete.
mcp: claude-praetorian-mcp (storage at .claude/praetorian/)
<br clear="left"/>
commands:
/compact-praetorian· save insights after research, decisions, or subagent work/restore-praetorian· load relevant past compactions at session start
tools:
praetorian_compact· save structured context: research, decisions, flow analysispraetorian_restore· search or list recent compactions by query or type
hooks:
PreToolUseonEnterPlanMode· lists prior compactions before planningPreCompacton*· saves context before compaction resets itPostToolUseonWebFetch/WebSearch· prompts to compact research findingsSubagentStopon*· prompts to compact subagent results
<h3>historian</h3>
<a href="https://github.com/Vvkmnn/claude-historian-mcp"><img src="logo/claude-historian.svg" width="200" height="200" align="left" alt="historian"></a>
Session searcher. Checks past sessions before you do redundant research, planning, or debugging to find solutions or insights we can recycle.
mcp: claude-historian-mcp (9 search tools across conversations, errors, files, plans)
<br clear="left"/>
commands:
/search-historian· search past sessions for solutions, patterns, or context
tools:
search_conversations· search conversation history, .claude files, and tasksfind_file_context· find all conversations related to a specific filefind_similar_queries· find previous similar questions with enhanced matchingget_error_solutions· find solutions for specific errorslist_recent_sessions· browse recent sessions with activity summariesextract_compact_summary· get intelligent summary of a session with key insightsfind_tool_patterns· analyze tool usage patterns and workflowssearch_plans· search plan files for past approaches and decisionssearch_config· search .claude config files for guidance and patterns
hooks:
PreToolUseonWebSearch/WebFetch· checks history before web researchPreToolUseonEnterPlanMode· searches past plans before planningPreToolUseonTask· checks tool patterns before launching agentsPostToolUseonBash· suggests error solutions after failures
<h3>oracle</h3>
<a href="https://github.com/Vvkmnn/claude-oracle-mcp"><img src="logo/claude-oracle.svg" width="200" height="200" align="left" alt="oracle"></a>
Tool recommender. Searches 17 sources in parallel to find relevant skills, plugins, and MCP servers.
mcp: claude-oracle-mcp (smithery, glama, npm, github, awesome-mcp-servers, and more)
<br clear="left"/>
commands:
/search-oracle· search all sources for skills, plugins, or MCP servers
tools:
search· search by query with optional type and semantic filtersbrowse· browse by category or popularitysources· show available data sources and their status
hooks:
PreToolUseonEnterPlanMode· searches for relevant tools before planningPostToolUseonBash· searches for tools that solve errors
<h3>gladiator</h3>
<a href="https://github.com/Vvkmnn/claude-gladiator-mcp"><img src="logo/claude-gladiator.svg" width="200" height="200" align="left" alt="gladiator"></a>
Continuous learner. Observes tool failures and prompts reflection at session end to evolve rules, hooks, and skills incrementally.
mcp: claude-gladiator-mcp (storage at .claude/gladiator/)
<br clear="left"/>
commands:
/review-gladiator· review and cluster observations into actionable patterns
tools:
gladiator_observe· record a pattern worth learning from, deduplicates by hashgladiator_reflect· query and cluster observations with recommendations
hooks:
PostToolUseonBash/Edit/Write· observes failure patterns (silent on success)Stopon*· prompts reflection if unprocessed observations exist
<h3>vigil</h3>
<a href="https://github.com/Vvkmnn/claude-vigil-mcp"><img src="logo/claude-vigil.svg" width="200" height="200" align="left" alt="vigil"></a>
Quick saver. Saves checkpoints before dangerous operations, diffs changes, restores files safely to augment /rewind.
mcp: claude-vigil-mcp (storage at .claude/vigil/)
<br clear="left"/>
commands:
/save-vigil· create a named checkpoint of tracked files/restore-vigil· restore files from a previous checkpoint
tools:
vigil_save· save a named checkpoint with optional descriptionvigil_list· list checkpoints, drill into one for file detailsvigil_diff· diff current files against a checkpointvigil_restore· restore files from a checkpointvigil_delete· delete a checkpoint or all checkpoints
hooks:
PreToolUseonBash· auto-quicksaves before destructive commands (rm, mv, etc.)
<h3>orator</h3>
<a href="https://github.com/Vvkmnn/claude-orator-mcp"><img src="logo/claude-orator.svg" width="200" height="200" align="left" alt="orator"></a>
Prompt coach. Deterministically scores and restructures prompts using Anthropic best practices. No LLM calls, no network, in-memory only.
mcp: claude-orator-mcp (in-memory, zero storage)
<br clear="left"/>
commands:
/reprompt-orator· score and optimize a prompt with detailed feedback
tools:
orator_optimize· analyze and optimize a prompt, returns score metrics and applied techniques
hooks:
PreToolUseonTask· suggests optimization for under-specified subagent prompts
architecture
Each plugin is a thin wrapper. hooks inject prompts that trigger MCP tools at high-impact moments. plugins contain no business logic: they tell claude _when_ to search, save, or restore. the MCP servers handle the actual work.
claude-emporium/
├── .claude-plugin/
│ ├── marketplace.json plugin registry (6 plugins)
│ └── plugin.json root manifest
│
├── plugins/
│ ├── claude-praetorian/ context guard → praetorian-mcp
│ │ ├── .claude-plugin/plugin.json declares praetorian MCP server
│ │ ├── hooks/
│ │ │ ├── hooks.json hook event bindings
│ │ │ ├── post-research.js compact after WebSearch/WebFetch
│ │ │ ├── post-subagent.js compact after Task completes
│ │ │ ├── pre-compact.js remind to compact before context reset
│ │ │ └── pre-plan.js list prior compactions before planning
│ │ ├── commands/
│ │ │ ├── compact-praetorian.md /compact
│ │ │ └── restore-praetorian.md /compact-status
│ │ ├── skills/claude-praetorian/SKILL.md
│ │ └── README.md
│ │
│ ├── claude-historian/ session memory → historian-mcp
│ │ ├── .claude-plugin/plugin.json declares historian MCP server
│ │ ├── hooks/
│ │ │ ├── hooks.json
│ │ │ ├── pre-websearch.js check history before web research
│ │ │ ├── pre-planning.js search past plans before planning
│ │ │ ├── pre-task.js check tool patterns before agents
│ │ │ └── post-error.js suggest past error solutions
│ │ ├── commands/search-historian.md /history
│ │ ├── skills/claude-historian/SKILL.md
│ │ └── README.md
│ │
│ ├── claude-oracle/ tool discovery → oracle-mcp
│ │ ├── .claude-plugin/plugin.json declares oracle MCP server
│ │ ├── hooks/
│ │ │ ├── hooks.json
│ │ │ ├── pre-planning.js search registries before planning
│ │ │ └── post-error.js find tools that solve errors
│ │ ├── commands/search-oracle.md /discover
│ │ ├── skills/claude-oracle/SKILL.md
│ │ └── README.md
│ │
│ ├── claude-gladiator/ learn & adapt → gladiator-mcp
│ │ ├── .claude-plugin/plugin.json declares gladiator MCP server
│ │ ├── hooks/
│ │ │ ├── hooks.json
│ │ │ ├── stop.js prompt reflection at session end
│ │ │ └── post-error.js observe failure patterns
│ │ ├── commands/review-gladiator.md /observe · /reflect
│ │ ├── skills/claude-gladiator/SKILL.md
│ │ └── README.md
│ │
│ ├── claude-vigil/ file recovery → vigil-mcp
│ │ ├── .claude-plugin/plugin.json declares vigil MCP server
│ │ ├── hooks/
│ │ │ ├── hooks.json
│ │ │ └── pre-bash.js quicksave before rm/reset/checkout
│ │ ├── commands/
│ │ │ ├── save-vigil.md /save
│ │ │ └── restore-vigil.md /restore · /snapshots
│ │ ├── skills/claude-vigil/SKILL.md
│ │ └── README.md
│ │
│ ├── claude-orator/ prompt rhetoric → orator-mcp
│ │ ├── .claude-plugin/plugin.json declares orator MCP server
│ │ ├── hooks/
│ │ │ ├── hooks.json
│ │ │ └── pre-task.js optimize prompts before subagents
│ │ ├── commands/reprompt-orator.md /optimize
│ │ ├── skills/claude-orator/SKILL.md
│ │ └── README.md
│ │
│ └── shared/
│ └── utils.js loadSettings · siblings · readStdin · emit
│
├── mcp/
│ └── .mcp.json all 6 MCP servers (standalone install)
│
├── logo/ plugin SVG icons
├── .github/ issue templates · PR template · CI workflows
├── scripts/readme.py README generation
└── README.md
each plugin.json declares its MCP server:
{ "mcpServers": { "praetorian": { "command": "npx", "args": ["claude-praetorian-mcp"] } } }
each hook script follows the same pattern:
readStdin() → loadSettings() → siblings() → emit(<system-reminder>)
runtime storage:
project/.claude/praetorian/ context snapshots
project/.claude/gladiator/ observations + reflections
project/.claude/vigil/ file checkpoints
~/.claude/settings.json enabledPlugins (auto) · claude-emporium (manual)zero overhead, by design:
- no external API calls. no network requests from hooks or plugins
- no background processes. hooks run inline and exit (~1ms each)
- no databases. praetorian, gladiator, and vigil store to project-local dirs (
.claude/). oracle, historian, and orator compute in-memory with TTL caching - no dependencies beyond Node.js (ships with Claude Code)
- all hook scripts are pure Node.js, single-purpose, and stateless
- built for claude, by claude
configuration
Sibling detection: /plugin install adds each plugin to enabledPlugins in ~/.claude/settings.json. Hooks read this to detect siblings and coordinate automatically. For example, historian's pre-research hook will note "praetorian will compact this" when both are installed, and oracle's error hook will mention "historian can search past solutions" if historian is present. This is what sibling detection looks like:
// ~/.claude/settings.json
{
// ...model, permissions, apiKey, etc.
"enabledPlugins": {
// ...other installed plugins
"claude-praetorian@claude-emporium": true, // detected as sibling by all other plugins
"claude-historian@claude-emporium": true, // detected as sibling by all other plugins
"claude-oracle@claude-emporium": true,
// ...added automatically by /plugin install
},
// ...
}Hook settings: every hook checks a boolean setting before running. All default to true. To disable any hook, add overrides under the claude-emporium key in ~/.claude/settings.json:
// ~/.claude/settings.json
{
"claude-emporium": {
"suggest_siblings": false, // stop suggesting uninstalled emporium plugins
"claude-praetorian": {
"auto_compact_research": false, // don't prompt to compact after WebFetch/WebSearch
"auto_compact_subagent": false, // don't prompt to compact after subagent completes
"check_compactions_before_plan": false, // don't list prior compactions before planning
"remind_compact": false, // don't remind to compact before context reset
},
"claude-historian": {
"search_before_web": false, // don't check history before web research
"search_before_plan": false, // don't search past plans before planning
"search_before_task": false, // don't check tool patterns before launching agents
"search_after_error": false, // don't suggest error solutions after failures
},
"claude-oracle": {
"search_before_plan": false, // don't search for relevant tools before planning
"search_after_error": false, // don't search for tools that solve errors
},
"claude-gladiator": {
"observe_after_failure": false, // don't record failure patterns
"reflect_before_stop": false, // don't prompt reflection at session end
},
"claude-vigil": {
"auto_quicksave": false, // don't quicksave before destructive commands
},
"claude-orator": {
"optimize_subagent_prompts": false, // don't suggest prompt optimization before Task
},
},
}Only include the settings you want to change. Omitted keys keep their defaults (true). The full defaults are defined in shared/utils.js.
synergy
Each plugin works standalone. when multiple are installed, they detect siblings at runtime and coordinate, usually without any extra configuration.
<!-- ASCII snapshot (closest WIP — alignment breaks in monospace due to emoji width):
VIGIL ──────────[5]──────────── PRAETORIAN ──────────[1]──────────── HISTORIAN file recovery context guard session memory | | | \ / | | | [2] [9] [7] [8] | [10] | | | \ / | | [11] ORACLE \ --------+------- [4] [6] | tool discovery\ | | | | | | --------+---|-------- | | | [3] [10] | | | | | | | | | | | | | +------------ ORATOR --------+----+-------- GLADIATOR ------+--+ | prompt rhetoric learn & adapt |
[1] praetorian <> historian past plans & context shared [2] praetorian <> oracle oracle discoveries in compactions [3] oracle <> gladiator tools found for error patterns [4] historian <> gladiator past solutions enrich reflection [5] vigil <> praetorian quicksave + context saved together [6] historian <> oracle past solutions + tool search [7] praetorian <> historian web search findings compacted [8] orator <> historian past well-scored prompts as examples [9] orator <> praetorian compact optimized results [10] orator <> gladiator observe which techniques improve scores [11] orator <> vigil checkpoint before batch rewrites [12] orator <> oracle find prompt engineering tools -->
%%{init: {'theme': 'base', 'htmlLabels': true, 'themeVariables': {'primaryColor': '#f5f5f5', 'primaryTextColor': '#1a1a1a', 'primaryBorderColor': '#cccccc', 'lineColor': '#555555', 'edgeLabelBackground': '#f0f0f0', 'fontSize': '16px'}, 'flowchart': {'nodeSpacing': 80, 'rankSpacing': 70, 'curve': 'basis', 'padding': 20}}}%%
flowchart TD
P["⚜️<br/>PRAETORIAN<br/>context guard"]:::gold
H["📜<br/>HISTORIAN<br/>session memory"]:::scroll
V["🏺<br/>VIGIL<br/>file recovery"]:::amphora
O["🔮<br/>ORACLE<br/>tool discovery"]:::crystal
G["⚔️<br/>GLADIATOR<br/>learn & adapt"]:::steel
R["🪶<br/>ORATOR<br/>prompt rhetoric"]:::quill
P ----|" plans & context "| H
V ----|" quicksave "| P
P ----|" compactions "| O
H ----|" solutions + tools "| O
O ----|" error tools "| G
H ----|" past solutions "| G
R ----|" scored prompts "| H
R ----|" compact results "| P
R ----|" find tools "| O
R ----|" observe techniques "| G
R ----|" checkpoint "| V
classDef gold fill:#fff8e1,stroke:#f9a825,stroke-width:2px,color:#5d4037
classDef scroll fill:#fff3e0,stroke:#ef6c00,stroke-width:2px,color:#4e342e
classDef amphora fill:#fbe9e7,stroke:#e64a19,stroke-width:2px,color:#bf360c
classDef crystal fill:#f3e5f5,stroke:#8e24aa,stroke-width:2px,color:#4a148c
classDef steel fill:#eceff1,stroke:#546e7a,stroke-width:2px,color:#263238
classDef quill fill:#efebe9,stroke:#795548,stroke-width:2px,color:#3e2723Enhanced behaviors when siblings are detected:
| event | plugin | alone | with siblings |
|---|---|---|---|
| plan | ⚜️ praetorian | lists prior compactions | + 📜 past plans, 🔮 finds tools |
| plan | 📜 historian | searches past plans | + 🔮 also searches for tools |
| plan | 🔮 oracle | searches 17 registries | + 📜 past plans, ⚜️ compactions |
| compact | ⚜️ praetorian | snapshots context | + 🔮 discoveries included |
| error | 📜 historian | searches past solutions | + 🔮 finds tools for this error class |
| error | 🔮 oracle | searches for error-solving tools | + 📜 checks past solutions |
| error | ⚔️ gladiator | observes failure pattern | + 📜 enriches with past solutions |
| research | 📜 historian | checks history first | + ⚜️ compacts findings after |
| subagent | ⚜️ praetorian | prompts to compact results | + ⚔️ observes subagent outcome patterns |
| task | 📜 historian | checks tool patterns | + 🔮 suggests tools for the task |
| task | 🪶 orator | optimizes vague prompts | + 📜 searches past scored prompts |
| stop | ⚔️ gladiator | prompts reflection | + 📜 and 🔮 deepen analysis |
| bash | 🏺 vigil | quicksaves before destructive | + ⚜️ preserves context alongside |
Detection is automatic via one fs.readFileSync call (~1ms), and falls back gracefully to solo mode if settings are missing or siblings are not dedicated.
contributing
Bug fixes, improvements, and documentation PRs are welcome, as are inspirations for new plugins and MCPs that could improve synergy and coverage. See CONTRIBUTING.md for full guidelines.
- Issues: use the issue templates. Bug reports, feature requests, and plugin ideas each have their own form, but please check open issues for duplicates first.
- PRs: Fork, branch, and follow the PR template. Good contributions include better hook triggers, bug fixes, documentation, and performance improvements, and any novelties that improve experience or perforamnce.
- New plugins: Please Open a plugin idea issue; always looking for plugins that have clear trigger events, often wrap specific MCPs, or provide obvious synergy with their sibling plugins. Above all, always looking to solve a concrete user pain for other Claude users.
troubleshooting
- plugin not triggering? Verify the mcp server is running:
claude mcp list
- mcp not found? Install via npm:
claude mcp add <name> -- npx <package>
- hook errors? Check that node.js is available and
CLAUDE_PLUGIN_ROOTis set.
license
MIT
<p align="center"> <img src="logo/consummation.jpg" width="100%" alt="The Consummation of Empire — Thomas Cole, 1836"> <br> <sub><a href="https://en.wikipedia.org/wiki/The_Course_of_Empire_(paintings)"><b>The Consummation of Empire</b></a> — <a href="https://en.wikipedia.org/wiki/Thomas_Cole"><b>Thomas Cole</b></a>, 1836</sub> <br><br> <em>"alea iacta est" | the die is cast</em> <br> <sub>— <a href="https://en.wikipedia.org/wiki/Julius_Caesar">Julius Caesar</a>, crossing the Rubicon</sub> </p>