T09 · Insecure Skill Coding Practices
Warning
- Location
summarize.sh:19- Finding
Untrusted Transcript Content Can Manipulate Model Output
- Content
View full analysis
- Remediation
View remediation
Security audit
Meeting Notes Summarizer
Security checks for vulnerabilities and agentic risk
Overview
The skill appears to be a straightforward meeting summarizer, but it sends full transcripts to Anthropic without clear user-facing disclosure.
Install only if you are comfortable sending any transcript you pipe into the script to Anthropic. Do not use it for regulated, confidential, HR, legal, customer, or credential-containing meetings unless that provider use is approved, and verify important decisions or action items against the original transcript.
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 (1)
Vulnerability Patterns
- Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
- Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
- Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (9)
Ae1
High
- Category
- analysis-evasion
- Confidence
- 100% confidence
- Finding
Referenced artifact was not completely inspected
- Content
md ./summarize.sh < transcript.txt
Ae1
High
- Category
- analysis-evasion
- Confidence
- 100% confidence
- Finding
Referenced artifact was not completely inspected
- Content
md ./summarize.sh < transcript.txt
Ae1
High
- Category
- analysis-evasion
- Confidence
- 100% confidence
- Finding
Referenced artifact was not completely inspected
- Content
md ./summarize.sh < transcript.txt
External Script Fetching
High
- Category
- Supply Chain
- Confidence
- 90% confidence
- Finding
Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.
- Content
sh print(json.dumps(prompt + transcript)) ") RESPONSE=$(curl -s https://api.anthropic.com/v1/messages \ -H "content-type: application/json" \ -H "x-api-key: $ANTHROPIC_API_KEY" \ -H "anthropic-version: 2023-06-01" \
Missing User Warnings
Medium
- Category
- Not specified by scanner
- Confidence
- 97% confidence
- Finding
The usage and requirements sections show that transcripts are processed using
curland anANTHROPIC_API_KEY, but they do not explicitly warn users that potentially sensitive meeting content will be transmitted to an external third-party API. Meeting transcripts commonly contain confidential business discussions, PII, credentials, or legal/HR information, so omission of this disclosure can lead to unintended data exfiltration and policy or compliance violations.- Content
Context-Inappropriate Capability
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The script requires an API key and transmits the full meeting transcript to Anthropic, which is a data exfiltration pathway by design. In a summarization skill this may be functionally intended, but without metadata, manifest scope, or explicit disclosure/consent, sensitive meeting content could be sent off-host unexpectedly.
- Content
External Transmission
Medium
- Category
- Data Exfiltration
- Confidence
- 93% confidence
- Finding
The hardcoded remote API endpoint confirms that processing occurs off-device and outside the user's immediate environment. For meeting transcripts, this can expose sensitive discussions to third-party handling and retention policies if users are not informed.
- Content
sh print(json.dumps(prompt + transcript)) ") RESPONSE=$(curl -s https://api.anthropic.com/v1/messages \ -H "content-type: application/json" \ -H "x-api-key: $ANTHROPIC_API_KEY" \ -H "anthropic-version: 2023-06-01" \
External Transmission
Medium
- Category
- Data Exfiltration
- Confidence
- 93% confidence
- Finding
The hardcoded remote API endpoint confirms that processing occurs off-device and outside the user's immediate environment. For meeting transcripts, this can expose sensitive discussions to third-party handling and retention policies if users are not informed.
- Content
sh print(json.dumps(prompt + transcript)) ") RESPONSE=$(curl -s https://api.anthropic.com/v1/messages \ -H "content-type: application/json" \ -H "x-api-key: $ANTHROPIC_API_KEY" \ -H "anthropic-version: 2023-06-01" \
Missing User Warnings
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The script silently uploads whatever is provided on stdin to an external LLM API, and users may reasonably assume a local summarizer from the interface alone. This creates privacy and compliance risk if transcripts contain confidential business, legal, HR, or customer data.
- Content