io.github.SaravananJaichandar/world-model-mcp

What we checked

Six static checks, weighted by risk. Every result reflects only what could be observed in the published package and repository — never intent.

• Injection surface

  unscannable25% of grade

  Tool descriptions/manifest scanned for instruction-injection patterns (imperatives at the model, hidden text, 'ignore previous', data-exfil URLs).

  No tool descriptions, server instructions, prompts, or resources were fetched; nothing to scan for injection.

• Supply chain

  fail25% of grade

  Package provenance: namespace verification, repo linkage, maintainer count, account age, postinstall scripts, typosquat distance.

  Python build script executes at build time: network call in build script

• Credential hygiene

  fail15% of grade

  How the server takes secrets (env vs plaintext config vs hardcoded); secrets appearing in tool schemas.

  world_model_server/_embedded_token.py: contains a hardcoded GitHub fine-grained PAT (gith…vv (len 93)). Reads secrets from the environment (e.g. adapters/pi/index.ts) — the recommended intake shape.

• Permission scope

  fail15% of grade

  Declared tools vs. breadth (filesystem, network, exec); flags shell-exec and unbounded filesystem access.

  HIGH: world_model_server/cli.py: references a sensitive credential path (/.cursor/) — the tool-poisoning exfiltration target. HIGH: shell/exec capability AND network egress together — download-and-execute risk (adapters/pi/index.ts: imports child_process and calls exec/spawn; tests/test_v072_http_transport.py: outbound HTTP/socket call). HIGH: shell/exec capability AND destructive filesystem writes together (adapters/pi/index.ts: imports child_process and calls exec/spawn; benchmarks/repeat-mistake/clone_repo.py: destructive filesystem call (shutil.rmtree/os.remove/etc.)). discloses shell/exec capability: adapters/pi/index.ts: imports child_process and calls exec/spawn discloses shell/exec capability: benchmarks/repeat-mistake/agent_runner.py: subprocess.* call discloses shell/exec capability: benchmarks/repeat-mistake/claude_client.py: subprocess.* call discloses shell/exec capability: benchmarks/repeat-mistake/clone_repo.py: subprocess.* call discloses shell/exec capability: benchmarks/repeat-mistake/score.py: subprocess.* call discloses shell/exec capability: examples/managed-agents-self-hosted/deploy_modal.py: subprocess.* call discloses shell/exec capability: hooks/src/world-model-inject.ts: imports child_process and calls exec/spawn discloses shell/exec capability: tests/test_v010_continue_features.py: subprocess.* call discloses shell/exec capability: tests/test_v010_hermes_features.py: subprocess.* call discloses shell/exec capability: tests/test_v010_openclaw_features.py: subprocess.* call discloses shell/exec capability: tests/test_v030_features.py: subprocess.* call discloses shell/exec capability: tests/test_v040_features.py: subprocess.* call discloses shell/exec capability: tests/test_v050_features.py: subprocess.* call discloses shell/exec capability: tests/test_v060_features.py: subprocess.* call discloses shell/exec capability: tests/test_v070_features.py: subprocess.* call discloses shell/exec capability: tests/test_v072_http_transport.py: subprocess.* call discloses shell/exec capability: tests/test_v073_features.py: subprocess.* call discloses shell/exec capability: tests/test_v074_features.py: subprocess.* call discloses shell/exec capability: tests/test_v075_features.py: subprocess.* call discloses shell/exec capability: tests/test_v076_features.py: subprocess.* call discloses shell/exec capability: tests/test_v080_features.py: subprocess.* call discloses shell/exec capability: world_model_server/cli.py: subprocess.* call discloses shell/exec capability: world_model_server/seeder.py: subprocess.* call discloses shell/exec capability: world_model_server/adapters/pi/index.ts: imports child_process and calls exec/spawn discloses shell/exec capability: world_model_server/hooks/world-model-inject.js: imports child_process discloses shell/exec capability: world_model_server/hooks/world-model-validate.js: imports child_process and calls exec/spawn discloses broad/destructive filesystem access: benchmarks/repeat-mistake/clone_repo.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) discloses broad/destructive filesystem access: benchmarks/repeat-mistake/failure_classifier.py: open(…, 'w'/'a'/'x') write-mode file access discloses broad/destructive filesystem access: benchmarks/repeat-mistake/learning_hook.py: open(…, 'w'/'a'/'x') write-mode file access discloses broad/destructive filesystem access: benchmarks/repeat-mistake/orchestrator.py: open(…, 'w'/'a'/'x') write-mode file access discloses broad/destructive filesystem access: benchmarks/repeat-mistake/score.py: open(…, 'w'/'a'/'x') write-mode file access discloses broad/destructive filesystem access: hooks/src/world-model-session.ts: destructive/writing fs.* call (write/unlink/rm/chmod/rename) discloses broad/destructive filesystem access: tests/test_ingest.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) discloses broad/destructive filesystem access: tests/test_knowledge_graph.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) discloses broad/destructive filesystem access: tests/test_pr_reviews.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) discloses broad/destructive filesystem access: tests/test_seeder.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) discloses broad/destructive filesystem access: tests/test_v030_features.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) discloses broad/destructive filesystem access: tests/test_v040_features.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) discloses broad/destructive filesystem access: tests/test_v050_features.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) discloses broad/destructive filesystem access: tests/test_v060_features.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) discloses broad/destructive filesystem access: tests/test_v070_features.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) discloses broad/destructive filesystem access: world_model_server/audit.py: open(…, 'w'/'a'/'x') write-mode file access discloses broad/destructive filesystem access: world_model_server/cli.py: open(…, 'w'/'a'/'x') write-mode file access discloses broad/destructive filesystem access: world_model_server/linters.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) discloses broad/destructive filesystem access: world_model_server/hooks/world-model-session.js: destructive/writing fs.* call (write/unlink/rm/chmod/rename) Capability DISCLOSURE, not a verdict: static analysis sees the primitive is present and reachable, not whether its use is attacker-controlled.

• Version behavior

  unscannable10% of grade

  Diff of tool definitions between versions; new permissions or changed descriptions in a patch release (the postmark-mcp class).

  no prior version to diff (first sight of io.github.SaravananJaichandar/[email protected]); version-drift / rug-pull cannot be evaluated. Cold-start risk is covered by point-in-time checks (injection_surface, supply_chain, credential_hygiene), not here.

• Transport config

  warn10% of grade

  Remote servers: TLS and auth mode (none/token/OAuth). Local servers: whether the manifest indicates it phones home.

  manifest/source references telemetry behavior — disclosure