io.github.ampres-ai/talamus

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

  pass25% of grade

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

  Python build script has no build-time network/exec; repository linkage verified

• Credential hygiene

  warn15% of grade

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

  tests/test_talamus_redact.py: embeds a PRIVATE KEY block. (in an example/sample file — lower severity, but a real key here still leaks.) Reads secrets from the environment (e.g. src/talamus/adapters/llm.py) — the recommended intake shape.

• Permission scope

  fail15% of grade

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

  HIGH: src/talamus/redact.py: references a sensitive credential path (id_rsa) — the tool-poisoning exfiltration target. HIGH: src/talamus/services/integrations.py: references a sensitive credential path (/.cursor/) — the tool-poisoning exfiltration target. HIGH: src/talamus/vault_import.py: references a sensitive credential path (/.ssh/) — the tool-poisoning exfiltration target. HIGH: tests/test_talamus_redact.py: references a sensitive credential path (id_rsa) — the tool-poisoning exfiltration target. HIGH: shell/exec capability AND network egress together — download-and-execute risk (src/talamus/adapters/llm.py: subprocess.* call; src/talamus/adapters/llm.py: outbound HTTP/socket call). HIGH: shell/exec capability AND destructive filesystem writes together (src/talamus/adapters/llm.py: subprocess.* call; src/talamus/ingest.py: path traversal ('../') combined with file access). discloses shell/exec capability: src/talamus/adapters/llm.py: subprocess.* call discloses shell/exec capability: src/talamus/bench.py: subprocess.* call discloses shell/exec capability: src/talamus/cli/lifecycle.py: subprocess.* call discloses shell/exec capability: src/talamus/scan.py: subprocess.* call discloses shell/exec capability: src/talamus/services/integrations.py: subprocess.* call discloses shell/exec capability: tests/test_benchmarks_one_screen.py: subprocess.* call discloses broad/destructive filesystem access: src/talamus/ingest.py: path traversal ('../') combined with file access discloses broad/destructive filesystem access: src/talamus/services/backup.py: open(…, 'w'/'a'/'x') write-mode file access discloses broad/destructive filesystem access: tests/test_talamus_cli.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) Capability DISCLOSURE, not a verdict: static analysis sees the primitive is present and reachable, not whether its use is attacker-controlled.

• Version behavior

  pass10% of grade

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

  tool surface, capabilities, install scripts, and publisher unchanged vs prior version 1.0.2 (patch bump); no rug-pull drift.

• 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 phone home behavior — disclosure