Back to skill

Security audit

axiom-hash-multi

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent local hashing tool, but its manifest verification and special-file handling can reach beyond the user’s intended files or run indefinitely on streams.

Review before installing if you may verify MANIFEST files from other people or automation. Use only trusted manifests, avoid running it on device files, FIFOs, or arbitrary symlinks, and prefer a patched version that confines manifest entries to an intended directory and rejects non-regular files by default.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T09 · Insecure Skill Coding Practices

Warning
Location
axiom_hash_multi.py:230
Finding

Manifest Path Traversal Enables Out-of-Scope File Hashing

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
axiom_hash_multi.py:93
Finding

Unbounded Hashing of Character Devices and Streams Enables Denial of Service

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (32)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

Although compare mode, manifest verification, and JSON output are not inherently unsafe, the finding also indicates interaction with special device files and subprocess/testing behavior beyond the declared scope. In a security review context, undisclosed parsing and verification features expand the attack surface, and special-file handling can lead to hangs, excessive reads, or unexpected behavior if not carefully constrained.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

Although compare mode, manifest verification, and JSON output are not inherently unsafe, the finding also indicates interaction with special device files and subprocess/testing behavior beyond the declared scope. In a security review context, undisclosed parsing and verification features expand the attack surface, and special-file handling can lead to hangs, excessive reads, or unexpected behavior if not carefully constrained.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill documentation and usage guidance are predominantly written in French, including operational instructions and feature descriptions, but there is no indication that the user can choose another language or that the skill is intentionally limited to a French-speaking context. This matches the policy category for language or locale constraints imposed without opt-in or justification.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
84% confidence
Finding

The skill documentation declares no explicit tool scope, yet the described and statically detected capabilities include file access and shell execution. In an agent setting, missing permission boundaries can cause the skill to run with broader ambient authority than intended, increasing the chance of unauthorized file operations or command execution if the implementation is invoked unsafely.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

This file contains natural-language instructions, usage examples, errors, and help text in French, which effectively forces a specific language for users. The policy allows locale constraints only when they are optional or clearly justified, and neither condition is present here.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest description presents the skill as a multi-algorithm hash generator for a file or string. However, the code and module documentation add a separate MANIFEST-processing feature that parses a manifest file and verifies hashes for many referenced files, which is broader than simple hashing of a direct input.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

Manifest verification accepts filenames from a MANIFEST and reads each referenced local file relative to the manifest directory, which expands file access beyond the single user-specified target implied by the skill purpose. An attacker who can supply or influence the MANIFEST can cause the tool to hash arbitrary readable local files, including via path traversal entries like '../secret', exposing existence and digest information for additional files.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill manifest describes a deterministic, local, zero-dependencies multi-hash generator, but this file imports and later uses subprocess execution to invoke the CLI itself and external tools like sha256sum, md5sum, and openssl. Spawning subprocesses and depending on host-installed binaries is not justified by the stated hashing purpose and expands capability beyond what users would expect from such a skill.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · stress_test.py (reported line 301)May include surrounding context.

python
print("=" * 70)

# F1: --algo invalid
result = subprocess.run(
    [sys.executable, "axiom_hash_multi.py", f1mb.name, "--algo", "sha999"],
    capture_output=True, text=True, cwd=Path(__file__).parent
)

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · stress_test.py (reported line 308)May include surrounding context.

python
report("F", "invalid --algo (sha999)", result.returncode == 1)

# F2: --all with file
result = subprocess.run(
    [sys.executable, "axiom_hash_multi.py", f1mb.name, "--all"],
    capture_output=True, text=True, cwd=Path(__file__).parent
)

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · stress_test.py (reported line 316)May include surrounding context.

python
report("F", "--all returns 5 lines", len(lines) == 5, f"got {len(lines)} lines")

# F3: --string with empty
result = subprocess.run(
    [sys.executable, "axiom_hash_multi.py", "", "--string"],
    capture_output=True, text=True, cwd=Path(__file__).parent
)

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · stress_test.py (reported line 325)May include surrounding context.

python
report("F", "empty --string", got == expected, f"got={got[:16]}..." if got != expected else "")

# F4: --string with unicode
result = subprocess.run(
    [sys.executable, "axiom_hash_multi.py", "🐺", "--string"],
    capture_output=True, text=True, cwd=Path(__file__).parent
)

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · stress_test.py (reported line 334)May include surrounding context.

python
report("F", "--string with emoji", got == expected, f"got={got[:16]}..." if got != expected else "")

# F5: stdin with no data
result = subprocess.run(
    [sys.executable, "axiom_hash_multi.py", "--stdin", "--algo", "sha256"],
    input="", capture_output=True, text=True, cwd=Path(__file__).parent
)

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · stress_test.py (reported line 343)May include surrounding context.

python
report("F", "stdin with no data", got == expected)

# F6: file doesn't exist
result = subprocess.run(
    [sys.executable, "axiom_hash_multi.py", "/tmp/this_does_not_exist_xyz.txt"],
    capture_output=True, text=True, cwd=Path(__file__).parent
)

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · stress_test.py (reported line 350)May include surrounding context.

python
report("F", "nonexistent file rejected", result.returncode == 1)

# F7: input is a directory
result = subprocess.run(
    [sys.executable, "axiom_hash_multi.py", str(tmpdir)],
    capture_output=True, text=True, cwd=Path(__file__).parent
)

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The manifest explicitly markets the skill as having 'zero dependencies,' but this test file validates behavior by calling sha256sum, md5sum, and openssl from the host environment. That behavior is semantically inconsistent with the advertised zero-dependency posture, because successful execution of these paths depends on external executables being installed.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · stress_test.py (reported line 392)May include surrounding context.

python
print("=" * 70)

# H1: Compare with sha256sum
sha256sum_result = subprocess.run(
    ["sha256sum", str(f1mb)], capture_output=True, text=True
)
if sha256sum_result.returncode == 0:

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · stress_test.py (reported line 404)May include surrounding context.

python
report("H", "sha256sum not available", True, "(skipped)")

# H2: Compare with md5sum
md5sum_result = subprocess.run(
    ["md5sum", str(f1mb)], capture_output=True, text=True
)
if md5sum_result.returncode == 0:

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · stress_test.py (reported line 416)May include surrounding context.

python
report("H", "md5sum not available", True, "(skipped)")

# H3: Compare with openssl sha512
openssl_result = subprocess.run(
    ["openssl", "dgst", "-sha512", str(f1mb)], capture_output=True, text=True
)
if openssl_result.returncode == 0:

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · test_v110_new_features.py (reported line 191)May include surrounding context.

python
def test_10_json_single(self):
        """--json doit retourner un JSON valide."""
        result = subprocess.run(
            [sys.executable, "axiom_hash_multi.py", str(self.f), "--algo", "sha256", "--json"],
            capture_output=True, text=True, cwd=Path(__file__).parent
        )

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · test_v110_new_features.py (reported line 201)May include surrounding context.

python
def test_11_json_all(self):
        """--all --json doit retourner les 5 algos en JSON."""
        result = subprocess.run(
            [sys.executable, "axiom_hash_multi.py", str(self.f), "--all", "--json"],
            capture_output=True, text=True, cwd=Path(__file__).parent
        )

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · test_v110_new_features.py (reported line 226)May include surrounding context.

python
def test_12_compare_match(self):
        """--compare avec bon hash doit exit 0 et dire MATCH."""
        result = subprocess.run(
            [sys.executable, "axiom_hash_multi.py", str(self.f),
             "--compare", self.expected],
            capture_output=True, text=True, cwd=Path(__file__).parent

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · test_v110_new_features.py (reported line 236)May include surrounding context.

python
def test_13_compare_mismatch(self):
        """--compare avec mauvais hash doit exit 1 et dire MISMATCH."""
        result = subprocess.run(
            [sys.executable, "axiom_hash_multi.py", str(self.f),
             "--compare", "0" * 64],
            capture_output=True, text=True, cwd=Path(__file__).parent

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · test_v110_new_features.py (reported line 246)May include surrounding context.

python
def test_14_compare_with_json(self):
        """--compare --json doit retourner un JSON structuré."""
        result = subprocess.run(
            [sys.executable, "axiom_hash_multi.py", str(self.f),
             "--compare", self.expected, "--json"],
            capture_output=True, text=True, cwd=Path(__file__).parent

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · test_v110_new_features.py (reported line 273)May include surrounding context.

python
def test_15_manifest_verify_ok(self):
        """--verify-manifest doit exit 0 si tout passe."""
        result = subprocess.run(
            [sys.executable, "axiom_hash_multi.py",
             "--verify-manifest", str(self.manifest)],
            capture_output=True, text=True, cwd=Path(__file__).parent