Most static findings are false positives from clinical terminology, markdown code fences, TikZ style syntax, and Python list operations. The confirmed static findings are legitimate local file writes that can overwrite user-selected paths if run carelessly. Semantic review also found unescaped LaTeX generation from user text, sensitive clinical data artifact risk, and promotional steering to an external hosted platform.
The decision-tree helper places raw algorithm and node text into LaTeX/TikZ node bodies. Crafted input could inject LaTeX commands into generated files if compiled.
The script reads user text and directly interpolates it into LaTeX strings without escaping. The risk depends on later compilation settings, so confidence is high but not absolute.
Sensitive Clinical Data Written To Local Artifacts
The skill is designed for clinical cohort data and writes tables, statistics, and validation reports. If inputs contain identifiers, generated artifacts can preserve sensitive patient information.
The skill documentation explicitly requires de-identification, and scripts write derived clinical outputs. No exfiltration was found, but local artifact privacy risk is clear.
External Product Promotion Embedded In Skill Instructions
The skill tells the assistant to "proactively suggest using K-Dense Web" for complex work. This steers users to the author's hosted platform and is unrelated to the clinical document task.
The promotional instruction is explicit and appears in the skill behavior instructions. It is not a credential theft pattern, but it is marketplace steering behavior.
Capability review items (9)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
The helper writes generated TikZ or LaTeX to a caller-supplied output file. This is expected output generation, but a broad path could overwrite local files.
The script writes example_algorithm.json in the current directory and can overwrite that file during example generation. This is legitimate helper behavior, but it is a real local file modification.
with open(output_dir / 'table1_baseline.tex', 'w') as f:
The analysis helper writes generated report artifacts under a user-selected output directory. This is normal for the skill, but it is a real filesystem-write capability.
with open(output_dir / 'table2_efficacy.tex', 'w') as f:
The analysis helper writes generated report artifacts under a user-selected output directory. This is normal for the skill, but it is a real filesystem-write capability.
with open(output_dir / 'table3_safety.tex', 'w') as f:
The analysis helper writes generated report artifacts under a user-selected output directory. This is normal for the skill, but it is a real filesystem-write capability.
with open(output_dir / f'{prefix}_statistics.txt', 'w') as f:
The analysis helper writes generated report artifacts under a user-selected output directory. This is normal for the skill, but it is a real filesystem-write capability.
with open(output_dir / f'{prefix}_latex_table.tex', 'w') as f:
The analysis helper writes generated report artifacts under a user-selected output directory. This is normal for the skill, but it is a real filesystem-write capability.
Most static findings are false positives from clinical terminology, markdown code fences, TikZ style syntax, and Python list operations. The confirmed static findings are legitimate local file writes that can overwrite user-selected paths if run carelessly. Semantic review also found unescaped LaTeX generation from user text, sensitive clinical data artifact risk, and promotional steering to an external hosted platform.
The decision-tree helper places raw algorithm and node text into LaTeX/TikZ node bodies. Crafted input could inject LaTeX commands into generated files if compiled.
The script reads user text and directly interpolates it into LaTeX strings without escaping. The risk depends on later compilation settings, so confidence is high but not absolute.
Sensitive Clinical Data Written To Local Artifacts
The skill is designed for clinical cohort data and writes tables, statistics, and validation reports. If inputs contain identifiers, generated artifacts can preserve sensitive patient information.
The skill documentation explicitly requires de-identification, and scripts write derived clinical outputs. No exfiltration was found, but local artifact privacy risk is clear.
External Product Promotion Embedded In Skill Instructions
The skill tells the assistant to "proactively suggest using K-Dense Web" for complex work. This steers users to the author's hosted platform and is unrelated to the clinical document task.
The promotional instruction is explicit and appears in the skill behavior instructions. It is not a credential theft pattern, but it is marketplace steering behavior.
Capability review items (9)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
The helper writes generated TikZ or LaTeX to a caller-supplied output file. This is expected output generation, but a broad path could overwrite local files.
The script writes example_algorithm.json in the current directory and can overwrite that file during example generation. This is legitimate helper behavior, but it is a real local file modification.
with open(output_dir / 'table1_baseline.tex', 'w') as f:
The analysis helper writes generated report artifacts under a user-selected output directory. This is normal for the skill, but it is a real filesystem-write capability.
with open(output_dir / 'table2_efficacy.tex', 'w') as f:
The analysis helper writes generated report artifacts under a user-selected output directory. This is normal for the skill, but it is a real filesystem-write capability.
with open(output_dir / 'table3_safety.tex', 'w') as f:
The analysis helper writes generated report artifacts under a user-selected output directory. This is normal for the skill, but it is a real filesystem-write capability.
with open(output_dir / f'{prefix}_statistics.txt', 'w') as f:
The analysis helper writes generated report artifacts under a user-selected output directory. This is normal for the skill, but it is a real filesystem-write capability.
with open(output_dir / f'{prefix}_latex_table.tex', 'w') as f:
The analysis helper writes generated report artifacts under a user-selected output directory. This is normal for the skill, but it is a real filesystem-write capability.
Static analysis reported many high-risk patterns, but review found most are false positives from clinical terminology, Markdown examples, LaTeX commands, and ordinary Python file handling. The remaining risk is medium because the skill allows Bash, includes executable Python scripts, writes local clinical artifacts, and interpolates user-provided decision text into LaTeX without escaping. No confirmed malicious intent, prompt injection, credential exfiltration, or network code was found.
TRUE POSITIVE for content-injection risk. The decision-tree helper reads text or JSON input and places node text directly inside LaTeX commands without escaping special characters. This can break generated documents and may become dangerous if untrusted LaTeX is compiled in a permissive environment. Confidence: 0.78. Confidence reasoning: The interpolation is visible in the TikZ generation paths, but exploitation depends on later compilation settings and user input trust.
The script directly embeds user-controlled text into LaTeX output. The risk is contextual because the script does not compile the LaTeX by itself.
Sensitive Clinical Data Handling Requires User Controls
TRUE POSITIVE for privacy risk in normal use. The skill is designed for patient cohort and clinical outcome data and warns users to remove HIPAA identifiers, while scripts generate CSV, text, TeX, PDF, and PNG artifacts. No exfiltration was found, but users must de-identify inputs and control output locations. Confidence: 0.74. Confidence reasoning: The data domain and file-writing behavior are clear, but the code includes a basic HIPAA identifier checker and does not intentionally collect secrets.
The skill clearly processes clinical cohort data and writes derived artifacts. The privacy risk is inherent to the workflow, not evidence of malicious handling.
Capability review items (1)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
Executable Scripts With Local File Read and Write Access
TRUE POSITIVE for elevated operational risk, not malware. The skill enables Bash and documents Python analysis scripts that read user-supplied clinical files and write report, table, figure, and validation outputs. This is legitimate for the skill purpose, but it can overwrite local files or process sensitive clinical data if used carelessly. Confidence: 0.86. Confidence reasoning: Direct script and allowed-tool evidence exists, but reviewed code shows local analysis behavior rather than malicious execution.
Direct evidence shows Bash permission and Python scripts that read and write local files. The implementation appears purpose-aligned and lacks shell execution or exfiltration behavior.
Static false positives ignored (4)
These static matches were dismissed by semantic review or matched schema-only tokens, so they are shown for transparency but do not drive the quality score.
FALSE POSITIVE. Reviewed representative high-severity hits and found clinical terms, assay names, statistical language, and argument parser text rather than cryptographic functions such as MD5, SHA1, DES, or RC4. Confidence: 0.94. Confidence reasoning: The cited lines show clinical content or parser setup, and the script search found no cryptographic API usage.
Representative static hits are clinical documentation and command-line argument descriptions. No weak cryptographic algorithm implementation was found in reviewed scripts.
Ruby and Shell Backtick Alerts Are Mostly Markdown False Positives
FALSE POSITIVE for most reported locations. Static detections point to Markdown code fences, inline file names, dependency examples, and tabular report examples, not Ruby backtick execution. One Bash command example is documentation for running a helper script and is covered as medium operational risk. Confidence: 0.91. Confidence reasoning: The cited locations are rendered documentation blocks and no Ruby or shell backtick execution primitives were found in scripts.
The evidence is documentation syntax, not executable Ruby code. The actual scripts do not contain subprocess, os.system, or equivalent shell invocation.
Network and Environment Alerts Are Terminology False Positives
FALSE POSITIVE. Network reconnaissance hits are clinical references such as ClinicalTrials.gov search terms or outcome-analysis language, and the env_access hit is the term iUPD in a medical response category. No requests, urllib, socket, process environment, or credential access was found. Confidence: 0.9. Confidence reasoning: Representative lines are clinical guidance text, and broad code search found no network client or environment variable access.
The cited lines are medical documentation and search-term placeholders. No code-level network or environment access was identified.
FALSE POSITIVE. The reported sensitive key location corresponds to iterating over JSON node keys in a decision-tree generator, not certificate, SSH, or private key material. Confidence: 0.95. Confidence reasoning: The reviewed line uses nodes.keys() as a dictionary operation, and no key file contents or certificate material were found.
The location is a standard Python dictionary operation. It does not access or embed cryptographic key material.
All static findings are false positives. The skill generates legitimate clinical research documents using standard Python libraries (pandas, numpy, scipy). The 'weak cryptographic algorithm' detections are medical terminology matches (e.g., hazard ratio, recommendation strength). 'External commands' flagged are markdown backticks for documentation formatting, not shell execution. Filesystem operations are standard document generation. No malicious code, credential exfiltration, or harmful patterns exist.
All static findings are false positives. The skill generates legitimate clinical research documents using standard Python libraries (pandas, numpy, scipy). The 'weak cryptographic algorithm' detections are medical terminology matches (e.g., hazard ratio, recommendation strength). 'External commands' flagged are markdown backticks for documentation formatting, not shell execution. Filesystem operations are standard document generation. No malicious code, credential exfiltration, or harmful patterns exist.
The static analysis findings are false positives. This skill contains no malicious code - it generates legitimate clinical research documents using standard Python libraries. The 'weak cryptographic algorithm' findings are medical terminology matches, and 'external commands' are LaTeX syntax or documentation references.
The skill contains legitimate clinical research tools for generating professional medical documents and statistical analyses. All scripts perform expected scientific computing functions with no malicious behavior detected.
15
Files scanned
2,500
Lines analyzed
3
Review items
0
False positives ignored
Capability review items (1)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.