io.github.sseshachala/conduct-cli

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.

  src/conduct_cli/main.py: assigns a literal to password (hard…re (len 21)) — non-provider secret literal; prefer environment intake. src/conduct_cli/main.py: assigns a literal to api_key (abc1…89 (len 18)) — non-provider secret literal; prefer environment intake. tests/test_guard_policy.py: assigns a literal to API_KEY (Abc1…89 (len 25)) — non-provider secret literal; prefer environment intake. Reads secrets from the environment (e.g. src/conduct_cli/guard.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/conduct_cli/guard.py: references a sensitive credential path (/.cursor/) — the tool-poisoning exfiltration target. HIGH: src/conduct_cli/paxel.py: references a sensitive credential path (/.cursor/) — the tool-poisoning exfiltration target. HIGH: tests/test_guard_policy.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/conduct_cli/guard.py: subprocess.* call; src/conduct_cli/api.py: outbound HTTP/socket call). HIGH: shell/exec capability AND destructive filesystem writes together (src/conduct_cli/guard.py: subprocess.* call; src/conduct_cli/main.py: destructive filesystem call (shutil.rmtree/os.remove/etc.)). discloses shell/exec capability: src/conduct_cli/guard.py: subprocess.* call discloses shell/exec capability: src/conduct_cli/guardmcp.py: subprocess.* call discloses shell/exec capability: src/conduct_cli/hook_precompact_template.py: subprocess.* call discloses shell/exec capability: src/conduct_cli/hook_session_start_template.py: subprocess.* call discloses shell/exec capability: src/conduct_cli/hook_stop_template.py: subprocess.* call discloses shell/exec capability: src/conduct_cli/hook_template.py: subprocess.* call discloses shell/exec capability: src/conduct_cli/main.py: subprocess.* call discloses shell/exec capability: src/conduct_cli/main.py: os.system/os.popen/os.exec* call discloses shell/exec capability: src/conduct_cli/paxel.py: subprocess.* call discloses dynamic code execution: src/conduct_cli/guard.py: uses eval/exec/__import__ discloses dynamic code execution: src/conduct_cli/hook_template.py: uses eval/exec/__import__ discloses dynamic code execution: src/conduct_cli/main.py: uses eval/exec/__import__ discloses broad/destructive filesystem access: src/conduct_cli/main.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) discloses broad/destructive filesystem access: src/conduct_cli/main.py: path traversal ('../') combined with file access 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.sseshachala/[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

  fail10% of grade

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

  source performs runtime self-update / download-and-execute (bypasses version_behavior); manifest/source references telemetry behavior — disclosure