Back to skill

Security audit

Tennis Grand Slam Trip Planner For Slam Chasers

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent tennis travel planner, but it asks users to install an unpinned global CLI and includes unsafe shared temporary-file handling.

Install only if you trust the flyai CLI publisher and are comfortable sending travel details to flyai. Prefer a pinned, project-local CLI install and avoid running the bundled script on shared machines unless its output directory is changed to a private per-run location.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (2)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:19
Finding

Unpinned Global Installation of a Third-Party CLI

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/plan-trip.sh:22
Finding

Predictable Shared Temporary Directory Allows Symlink Attacks and Result Tampering

Content
View full analysis
"$OUTPUT_DIR/flights.json" 2>&1; then ``` ```bash if flyai search-hotel --dest-name "$CITY" --poi-name "$VENUE" --check-in-date "$ARRIVE_DATE" --check-out-date "$LEAVE_DATE" --sort distance_asc --max-price "$BUDGET" > "$OUTPUT_DIR/hotels.json" 2>&1; then ``` ```bash if flyai search-hotel --dest-name "$CITY" --poi-name "$VENUE" --check-in-date "$ARRIVE_DATE" --check-out-date "$LEAVE_DATE" --sort distance_asc > "$OUTPUT_DIR/hotels.json" 2>&1; then ``` ```bash YEAR=$(echo "$ARRIVE_DATE" | cut -d'-' -f1) flyai keyword-search --query "${SLAM_NAME}门票 ${YEAR}" > "$OUTPUT_DIR/tickets.json" 2>&1 || true flyai keyword-search --query "${CITY} 网球体验" > "$OUTPUT_DIR/experiences.json" 2>&1 || true flyai keyword-search --query "${COUNTRY}签证" > "$OUTPUT_DIR/visa.json" 2>&1 || true ``` ```bash flyai search-poi --city-name "$CITY" --category "城市观光" > "$OUTPUT_DIR/poi_urban.json" 2>&1 || true flyai search-poi --city-name "$CITY" --category "博物馆" > "$OUTPUT_DIR/poi_museum.json" 2>&1 || true flyai search-poi --city-name "$CITY" --category "人文古迹" > "$OUTPUT_DIR/poi_culture.json" 2>&1 || true ``` ### Technical Analysis The script writes all results to the fixed path `/tmp/slam-trip-results` using predictable filenames. It creates the directory with `mkdir -p` but does not verify its owner, permissions, file type, or whether existing result files are symbolic links. On a multi-user system, another local process may pre-create the directory or predictable result paths. Shell output redirection follows symbolic links. Consequently, a pre-created path such as `/tmp ...[truncated 2356 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (14)

Direct Prompt Extraction

High
Category
System Prompt Leakage
Confidence
85% confidence
Finding

Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Content

Scanner excerpt · SKILL.md (reported line 113)May include surrounding context.

md
6. Practical tips (visa, weather, transport, etiquette)
7. Source attribution: "Based on fly.ai real-time results"

## Output rules

- All output in valid Markdown
- Hotel images: `![hotel]({mainPic})`

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill directs the agent to collect travel details such as departure city and trip dates, then transmit them to external flyai APIs, but it provides no instruction to disclose that sharing to the user or obtain consent first. This creates a privacy and transparency issue because sensitive itinerary information may be sent off-platform without the user understanding where their data is going.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This markdown template forces a specific language/locale in its user-facing content, including headings and itinerary text, with no indication that users can opt into another language. That is a natural-language policy concern under the locale-choice rule because the file appears intended for end-user output.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

Several parameter descriptions require city names in Chinese, and the examples throughout the file are exclusively Chinese-language inputs. This creates a locale/language constraint in the skill documentation without offering a user choice or explaining why the restriction is required.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The reference data includes flyai dest-name values in Chinese, which imposes a specific language representation in a shared data file. Because the file does not state that Chinese output is optional, user-selected, or required for a region-specific integration, this creates a natural-language locale policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The POI category labels are written in Chinese in a general-purpose reference markdown file, which can force a specific language/locale on downstream skill behavior. No nearby text explains that Chinese is required by an external API or that the user can choose another language.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This destination-name field is encoded in Chinese in a file otherwise written in English, signaling a forced locale without opt-in. The absence of justification or alternative locale handling makes this a policy-style natural-language issue.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

These category values are presented only in Chinese even though the rest of the file is English, which can constrain downstream outputs or API requests to a single locale. There is no documented user opt-in or regional requirement to justify the language restriction.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The flyai dest-name value is supplied only in Chinese, which is a language-specific policy choice in an otherwise English document. Without explanation or opt-in, this may cause the skill to force a language the user did not request.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The recommended category tokens are provided solely in Chinese and could be propagated into generated responses or searches. The file does not indicate that Chinese is mandatory for an external system or that users can choose a different language.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This field uses a Chinese city name inside a broadly applicable English reference file, which constitutes a language-specific constraint. Because no locale rationale or user choice is provided, it may violate language/locale policy expectations.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The category names are specified only in Chinese, which can steer downstream behavior toward a fixed locale. The markdown does not justify this as a region-specific requirement or offer any opt-in/out mechanism.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill is scoped to tennis event travel planning, but it also performs a visa-related search using the user's destination country. This is a scope expansion beyond the stated purpose and can cause the agent to collect or infer immigration-related information the user did not request, increasing privacy and data-minimization risk. In a travel-planning context this is somewhat plausible, but because the skill description explicitly says it is for Grand Slam itinerary generation rather than general travel/immigration assistance, the extra query is unjustified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The document repeatedly frames guidance specifically for 'Chinese passport holders' and uses Chinese-language search queries, which signals a fixed locale/audience assumption rather than a user-selected preference. Because this is natural-language guidance in a markdown file and no opt-in or audience limitation is stated, it can be interpreted as a locale-specific policy choice without user choice.

Content

No source excerpt is available for this finding.