godkiller-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| GODKILLER_TOOLS_DIR | No | Path to directory containing helper binaries (rg, fd, semgrep, ast-grep) if not on PATH |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| gk_routeA | Classify intent into /ask|/plan|/debug|/ultradeep|/view|/verify. Pass action= and optional args={}. 1 actions; unknown action returns allowed list (saves schema tokens). |
| gk_taskC | Task lifecycle: open, hypothesize, graph, policy, blast_radius, edit_safe, failing_slice. Pass action= and optional args={}. 7 actions; unknown action returns allowed list (saves schema tokens). |
| gk_phaseA | Phase machine: assert, claim_done, rubric. Blocks illegal Antigravity phase skips. Pass action= and optional args={}. 3 actions; unknown action returns allowed list (saves schema tokens). |
| gk_evidenceC | Evidence: submit, capture_shot, visual_critic, visual_step (~10-shot QA), visual_sequence, screenshot, journey, inspect_image. Pass action= and optional args={}. 8 actions; unknown action returns allowed list (saves schema tokens). |
| gk_verifyA | Verification: bundle, exit (short stage_board), soak, probe, loop_*, competitor, compare, ladder. Pass action= and optional args={}. 11 actions; unknown action returns allowed list (saves schema tokens). |
| gk_memoryC | Workflow memory graph: lessons, marathon, query_graph, what_blocked, upsert_episode. Pass action= and optional args={}. 12 actions; unknown action returns allowed list (saves schema tokens). |
| gk_codeB | Code intel helpers (map/search/read). council/swarm/pipeline/self_heal = best-effort, not magic fix or formal proof. Pass action= and optional args={}. 21 actions; unknown action returns allowed list (saves schema tokens). |
| gk_guardA | Write allowlist for host PreToolUse. set_paths arms ONE turn (ship default max 1 path); end_turn clears it. Ultradeep syncs current file only — never the whole queue. Blocks native Write only when the host hook is wired. Pass action= and optional args={}. 3 actions; unknown action returns allowed list (saves schema tokens). |
| gk_scanA | Best-effort regex CWE heuristics (signal, not a pro audit); optional semgrep CLI. Pass action= and optional args={}. 2 actions; unknown action returns allowed list (saves schema tokens). |
| gk_browserA | Playwright fallback ONLY when chrome-devtools is NOT on the host. If chrome-devtools is listed, returns prefer_chrome_devtools — use that peer first; force_gk_browser=1 to override. Pass action= and optional args={}. 7 actions; unknown action returns allowed list (saves schema tokens). |
| gk_modeB | Modes/protocols/skills + ultradeep/view + debug_engine (static scan + token heuristics — not a live debugger) + tool_propose (search≠install) + plan_refute wake. Pass action= and optional args={}. 29 actions; unknown action returns allowed list (saves schema tokens). |
| gk_handoffC | Spec/feedback handoff gates. Pass action= and optional args={}. 4 actions; unknown action returns allowed list (saves schema tokens). |
| gk_metaB | Honesty status (disk MCP configs + real facades) + secrets key listing + 9-step plan template/validate. Pass action= and optional args={}. 4 actions; unknown action returns allowed list (saves schema tokens). |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 13 tools
Each tool maps to a distinct functional domain (code, guard, scan, route, task, phase, evidence, verify, memory, browser, mode, handoff, meta). The action namespaces are clearly separated, so an agent can reliably select the right tool for a given task. No two tools appear to do the same thing.
All tools follow a consistent gk_<noun> pattern with lowercase snake_case. The shared prefix and simple noun suffixes make the hierarchy predictable and easy to remember.
With 13 tools, the server is well within the recommended range. The broad domains are logically partitioned, and the action subcommands are packed inside each tool without exploding the tool list.
The toolset covers the major phases of a coding workflow: code observation, security scanning, task tracking, verification, evidence, memory, browsing, mode management, and meta/honesty. No critical gaps are apparent for its stated purpose of aiding complex development processes.