Back to skill

Security audit

Info Vivid

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly does what it says, but its HTML chart generator can turn chart data into executable browser content, so it needs review before use.

Use this only with trusted chart/report data unless the SVG/HTML escaping and color validation are fixed. Avoid opening or publishing generated HTML from untrusted JSON, pin Pillow in an isolated environment for PNG rendering, and be aware that outputs include WorkBuddy branding and can be copied to an archive path when requested.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (3)

T01 · Skill Instruction Hijacking

Note
Location
scripts/svg_bar_chart.py:238
Finding

Unconditional Third-Party Branding in Generated Reports

Content
View full analysis
{footer} Powered by WorkBuddy ``` ```python main_text = f"Report time: {now} | {text}" tw = draw_obj.textbbox((0, 0), main_text, font=F_SMALL)[2] draw_obj.text(((self.W - tw) // 2, y0 + 12), main_text, font=F_SMALL, fill=GRAY) if next_update: nt_w = draw_obj.textbbox((0, 0), next_update, font=F_SMALL)[2] draw_obj.text(((self.W - nt_w) // 2, y0 + 34), next_update, font=F_SMALL, fill=DIVIDER) # WorkBuddy watermark wm = "Powered by WorkBuddy" wm_w = draw_obj.textbbox((0, 0), wm, font=F_SMALL)[2] draw_obj.text((self.W - wm_w - PAD, y0 + 22), wm, font=F_SMALL, fill=(38, 58, 82)) ``` ### Technical Analysis Both output generators insert the phrase `Powered by WorkBuddy` without consulting user configuration. The SVG renderer places it directly in every generated document, while the PNG renderer draws it into the final image. The Skill documentation also defines the watermark as part of the expected design. Because the attribution cannot be disabled, generated artifacts are systematically modified with third-party promotional content unrelated to the supplied visualization data. This may misrepresent the author, producer, or sponsor of a report. It is classified as instruction hij ...[truncated 798 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
scripts/svg_bar_chart.py:129
Finding

Stored Script Injection in Generated SVG and HTML

Content
View full analysis
= 0 else f"{val:.2f}{suffix}" tooltip = label if extra: tooltip += f" | {extra}" if tag: tooltip += f" | {tag}" tooltip += f" | {display_val}" rows_svg.append(f""" {i+1:02d}. {label} {tooltip} {display_val} {tag}""") ``` ```python subtitle_svg = ( f'
Remediation
View remediation
`. 2. Escape attribute values separately, including quotation marks. 3. Prefer constructing SVG with an XML library such as `xml.etree.ElementTree` rather than concatenating markup with f-strings. 4. Apply strict validation to configurable colors: - Accept only known color names or patterns such as `^#[0-9A-Fa-f]{6}$`. - Reject CSS functions, quotes, semicolons, and arbitrary attribute fragments. 5. Validate numeric layout settings: - Require finite integers or floats. - Enforce safe minimum and maximum dimensions. - Reject `NaN`, infinity, negative widths, and invalid reference values. 6. Escape the HTML document title independently from SVG content. 7. Consider applying a restrictive Content Security Policy to generated HTML, such as disallowing scripts and external resources. This should be defense in depth, not a substitute for encoding. 8. Add regression tests using payloads in every input-controlled field, including labels, titles, KPI values, tags, tooltips, footer text, reference labels, colors, and theme values. ]]>

T08 · Insecure Dependencies

Warning
Location
SKILL.md:25
Finding

Unpinned Third-Party Dependency Installation Guidance

Content
View full analysis
Remediation
View remediation
`. 2. Generate a lock file containing cryptographic hashes for all resolved distributions. 3. Install dependencies with hash enforcement, such as `pip install --require-hashes -r requirements.txt`. 4. Document the expected trusted package index and avoid untrusted or implicit mirrors. 5. Install dependencies in an isolated virtual environment with minimal permissions. 6. Use automated dependency scanning and update pinned versions through a controlled review process. 7. Prefer prebuilt, verified wheels where appropriate, and verify provenance or signatures when the package ecosystem supports them. ]]>
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (7)

Tp4

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

The documentation presents the skill as a visualization utility, but the described behavior includes file archival/copy operations and does not cleanly align with the claimed output modes. Description-behavior mismatches are dangerous because users and orchestration layers may trust a narrower capability set than the skill actually uses, enabling unexpected persistence of generated content or invocation under false assumptions.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding

The skill documents behavior that reads input files and writes output artifacts, but it declares no explicit tool scope or permissions boundary. In an agent environment, this weakens policy enforcement and can allow the skill to be invoked with broader file access than users expect, especially when combined with automatic output and archival behavior.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger phrases are broad enough to match many ordinary user requests, increasing the chance the skill activates in contexts where the user did not intend file generation or report rendering. Overbroad activation is risky because it can cause unexpected handling of user data and unintended file writes, especially in multi-skill agent systems.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The markdown describes writing output files and optionally copying them to an archive path, but it does not clearly warn users that local files will be created or duplicated. In an agent setting, silent or poorly signposted file writes can surprise users, leak sensitive rendered data into persistent storage, or violate least-surprise expectations.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The generated HTML sets lang="zh-CN", and the script’s user-facing strings and font choices are consistently Chinese-specific throughout the file. The policy allows locale constraints only when the user can opt in or when the constraint is clearly documented and justified, neither of which is present here.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The skill is described primarily as rendering structured data into SVG/HTML charts and PNG reports. SKILL.md additionally documents copying outputs into an archive directory such as C:/Users/user/.ai-memory/news / C:\Users\ToddC\.ai-memory\news\, which is not necessary to generate visualizations and adds a separate persistence/archival behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The manifest describes this skill as rendering charts and long-form reports, but the CLI also supports copying the generated output into an arbitrary archive directory. That file-management behavior is not part of the stated visualization/rendering scope and goes beyond producing the requested chart/report artifact.

Content

No source excerpt is available for this finding.