ClawHub
Back to skill

Security audit

Wjs Editing Multicam

Security checks for vulnerabilities and agentic risk

Overview

The skill is a local multicam video editor, but it needs Review because crafted edit-plan files can be fed into FFmpeg without enough validation.

Use this only with EDL files you generated yourself from trusted media, or after validating the JSON fields and bounds. Avoid rendering EDLs from other people in an unrestricted environment, and keep backups because FFmpeg runs with overwrite enabled for the chosen output path.

A.I.G

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
scripts/render_cuts.py:20
Finding

Unvalidated EDL Values Allow FFmpeg Filter-Graph Injection in the Hard-Cut Renderer

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/render_pip.py:84
Finding

Unvalidated EDL Values Allow FFmpeg Filter-Graph Injection in the PiP Renderer

Content
View full analysis
0: pip_chain += f"pad={pip_total_w}:{pip_total_h}:{bw}:{bw}:white," pip_chain += f"setsar=1,fps={args.fps}[p{i}]" filters.append(pip_chain) filters.append(f"[m{i}][p{i}]overlay={x_expr}:{y_expr}:eof_action=pass[v{i}]") concat = "".join(f"[v{i}]" for i in range(len(edl))) filters.append(f"{concat}concat=n={len(edl)}:v=1:a=0[vout]") audio_offset = edl[0]["start"] if edl else 0.0 dur = plan["duration_sec"] fc = ";".join(filters) fc += (f";[{audio_src}:a:0]atrim=start={audio_offset}:" f"duration={dur},asetpts=PTS-STARTPTS[aout]") ``` The explicit PiP override is also accepted from the EDL without range validation: ```pytho ...[truncated 2921 chars]
Remediation
View remediation

SkillSpector

Vulnerability Patterns
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (8)

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The core behavior broadly aligns with multicam auto-editing from synced recordings using audio energy and sidecar timing data. However, the declared description overstates the delivered functionality in material ways. This script performs analysis and edit decision generation only: it extracts audio, computes envelopes, aligns them by sidecar offsets, selects camera segments, and writes an EDL JSON file. There is no code here to render or mux a final combined MP4, and no picture-in-picture overlay implementation. Those are significant user-visible capability gaps rather than minor implementation details, so this should be flagged as a mismatch.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The declared description presents this skill as an end-to-end multicam auto-editing tool that ingests multiple synced recordings and combines them by automatically selecting cameras based on audio energy, optionally with PiP. The supplied code chunk only renders a previously generated edit decision list (EDL) into an MP4. It performs trimming, scaling, padding, concatenation, and audio trimming via ffmpeg. The docstring explicitly states 'Hard cuts only (no transitions / no PiP),' and there is no logic for analyzing audio energy or choosing camera switches. Therefore, the code's actual behavior is materially narrower and different from the declared primary purpose.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding

The skill describes capabilities that read and write files and invoke shell-based tooling like ffmpeg, but it does not declare any explicit tool scope or permissions boundary. In an agent environment, that omission can cause over-broad tool access, making it easier for the skill to operate on unintended files or execute commands beyond its stated purpose.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The documentation says each input must have a .sync.json sidecar, but the implementation silently falls back to delta=0 and full coverage when sidecars are missing or malformed. In this editing context, that can cause misaligned timelines and incorrect camera selection without failing fast, producing deceptive or materially wrong output from untrusted or incomplete inputs.

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 · scripts/autoedit.py (reported line 33)May include surrounding context.

python
SCHEMA_VERSION = 1

def extract(video: Path, dst: Path):
    subprocess.run(["ffmpeg", "-nostdin", "-y", "-i", str(video),
                    "-map", "0:a:0", "-ac", "1", "-ar", str(SR),
                    "-f", "s16le", str(dst)], check=True, stderr=subprocess.DEVNULL)

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 · scripts/render_cuts.py (reported line 69)May include surrounding context.

python
str(args.out),
    ])
    print(" ".join(cmd))
    subprocess.run(cmd, check=True)

if __name__ == "__main__":
    main()

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 · scripts/render_pip.py (reported line 130)May include surrounding context.

python
str(args.out),
    ])
    print(" ".join(cmd))
    subprocess.run(cmd, check=True)

if __name__ == "__main__":
    main()

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 · scripts/render_pip.py (reported line 188)May include surrounding context.

python
str(args.out),
    ])
    print(" ".join(cmd))
    subprocess.run(cmd, check=True)

if __name__ == "__main__":
    main()