Back to skill

Security audit

transcript

Security checks for vulnerabilities and agentic risk

Overview

The skill retrieves YouTube transcripts through a disclosed service, with user-directed account setup and credential storage.

TranscriptOut receives your requested video identifiers and uses your API credits. Prefer your runtime's secret store. Optional onboarding can create an account or issue a new key for an existing account. The fallback stores a local key and changes shell startup configuration; remove those settings and revoke the key when finished. Both artifact files were reviewed, resolving the scanner's reference gaps; authenticated server behavior and publisher identity were not independently verified.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (14)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
85% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · references/auth-setup.md (reported line 117)May include surrounding context.

Run this as one command, with the code in place of CODE:

bash
d=~/.config/transcriptout; t=$(sed -n 's/.*"access_token":"\([^"]*\)".*/\1/p' "$d/signup.json"); (umask 077; curl -s -X POST https://api.transcriptout.com/auth/verify-cli -H "Authorization: Bearer $t" -H "Content-Type: application/json" -H "X-Client: youtube-skills-clawhub/transcript" -d '{"otp":"CODE","source":"youtube-skills-clawhub/transcript"}' -o "$d/verify.json"); k=$(sed -n 's/.*"api_key":"\([^"]*\)".*/\1/p' "$d/verify.json"); if [ -n "$k" ]; then (umask 077; printf '%s' "$k" > "$d/key"); rm -f "$d/signup.json"; echo stored; fi; grep -o '"code":"[^"]*"\|"credits":[0-9]*' "$d/verify.json"; rm -f "$d/verify.json"

It prints stored and the credit balance on success, or an error code. The key goes from the

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
85% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · references/auth-setup.md (reported line 117)May include surrounding context.

Run this as one command, with the code in place of CODE:

bash
d=~/.config/transcriptout; t=$(sed -n 's/.*"access_token":"\([^"]*\)".*/\1/p' "$d/signup.json"); (umask 077; curl -s -X POST https://api.transcriptout.com/auth/verify-cli -H "Authorization: Bearer $t" -H "Content-Type: application/json" -H "X-Client: youtube-skills-clawhub/transcript" -d '{"otp":"CODE","source":"youtube-skills-clawhub/transcript"}' -o "$d/verify.json"); k=$(sed -n 's/.*"api_key":"\([^"]*\)".*/\1/p' "$d/verify.json"); if [ -n "$k" ]; then (umask 077; printf '%s' "$k" > "$d/key"); rm -f "$d/signup.json"; echo stored; fi; grep -o '"code":"[^"]*"\|"credits":[0-9]*' "$d/verify.json"; rm -f "$d/verify.json"

It prints stored and the credit balance on success, or an error code. The key goes from the

Chaining Abuse

High
Category
Tool Misuse
Confidence
75% confidence
Finding

Tool calls are chained to bypass individual safety checks or escalate capabilities beyond what any single tool call would allow.

Content

Scanner excerpt · references/auth-setup.md (reported line 117)May include surrounding context.

Run this as one command, with the code in place of CODE:

bash
d=~/.config/transcriptout; t=$(sed -n 's/.*"access_token":"\([^"]*\)".*/\1/p' "$d/signup.json"); (umask 077; curl -s -X POST https://api.transcriptout.com/auth/verify-cli -H "Authorization: Bearer $t" -H "Content-Type: application/json" -H "X-Client: youtube-skills-clawhub/transcript" -d '{"otp":"CODE","source":"youtube-skills-clawhub/transcript"}' -o "$d/verify.json"); k=$(sed -n 's/.*"api_key":"\([^"]*\)".*/\1/p' "$d/verify.json"); if [ -n "$k" ]; then (umask 077; printf '%s' "$k" > "$d/key"); rm -f "$d/signup.json"; echo stored; fi; grep -o '"code":"[^"]*"\|"credits":[0-9]*' "$d/verify.json"; rm -f "$d/verify.json"

It prints stored and the credit balance on success, or an error code. The key goes from the

Chaining Abuse

High
Category
Tool Misuse
Confidence
75% confidence
Finding

Tool calls are chained to bypass individual safety checks or escalate capabilities beyond what any single tool call would allow.

Content

Scanner excerpt · references/auth-setup.md (reported line 117)May include surrounding context.

Run this as one command, with the code in place of CODE:

bash
d=~/.config/transcriptout; t=$(sed -n 's/.*"access_token":"\([^"]*\)".*/\1/p' "$d/signup.json"); (umask 077; curl -s -X POST https://api.transcriptout.com/auth/verify-cli -H "Authorization: Bearer $t" -H "Content-Type: application/json" -H "X-Client: youtube-skills-clawhub/transcript" -d '{"otp":"CODE","source":"youtube-skills-clawhub/transcript"}' -o "$d/verify.json"); k=$(sed -n 's/.*"api_key":"\([^"]*\)".*/\1/p' "$d/verify.json"); if [ -n "$k" ]; then (umask 077; printf '%s' "$k" > "$d/key"); rm -f "$d/signup.json"; echo stored; fi; grep -o '"code":"[^"]*"\|"credits":[0-9]*' "$d/verify.json"; rm -f "$d/verify.json"

It prints stored and the credit balance on success, or an error code. The key goes from the

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 41)May include surrounding context.

Fetch the transcript of any YouTube video.

bash
curl -s "https://api.transcriptout.com/v1/transcript?video=VIDEO_URL_OR_ID&format=text" \
  -H "Authorization: Bearer $TRANSCRIPTOUT_API_KEY" \
  -H "X-Client: youtube-skills-clawhub/transcript"

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/auth-setup.md (reported line 172)May include surrounding context.

prints the account id and the credits, never the key:

bash
curl -s https://api.transcriptout.com/auth/me \
  -H "Authorization: Bearer $TRANSCRIPTOUT_API_KEY" \
  -H "X-Client: youtube-skills-clawhub/transcript"

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/auth-setup.md (reported line 61)May include surrounding context.

it and writes the private file:

bash
mkdir -p ~/.config/transcriptout && chmod 700 ~/.config/transcriptout && (umask 077; printf 'TranscriptOut key: '; stty -echo; IFS= read -r k; stty echo; echo; printf '%s' "$k" > ~/.config/transcriptout/key)

Then continue with Load the key below.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/auth-setup.md (reported line 91)May include surrounding context.

Run this as one command, with the user's address in place of USER_EMAIL:

bash
mkdir -p ~/.config/transcriptout && chmod 700 ~/.config/transcriptout && (umask 077; curl -s -X POST https://api.transcriptout.com/auth/register-cli -H "Content-Type: application/json" -H "X-Client: youtube-skills-clawhub/transcript" -d '{"email":"USER_EMAIL","source":"youtube-skills-clawhub/transcript"}' -o ~/.config/transcriptout/signup.json); grep -o '"ok":[a-z]*\|"code":"[^"]*"' ~/.config/transcriptout/signup.json

It prints only "ok":true, or "ok":false with an error code. The response itself holds

The skill performs account registration and API-key issuance despite explicitly excluding account chores.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The manifest says to skip this skill for account chores, but this workflow executes account registration and verification requests at L091 and L117. L133 also explicitly extends the workflow to signing into existing accounts and issuing new keys, which changes account state rather than retrieving or analyzing video transcripts.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/auth-setup.md (reported line 91)May include surrounding context.

Run this as one command, with the user's address in place of USER_EMAIL:

bash
mkdir -p ~/.config/transcriptout && chmod 700 ~/.config/transcriptout && (umask 077; curl -s -X POST https://api.transcriptout.com/auth/register-cli -H "Content-Type: application/json" -H "X-Client: youtube-skills-clawhub/transcript" -d '{"email":"USER_EMAIL","source":"youtube-skills-clawhub/transcript"}' -o ~/.config/transcriptout/signup.json); grep -o '"ok":[a-z]*\|"code":"[^"]*"' ~/.config/transcriptout/signup.json

It prints only "ok":true, or "ok":false with an error code. The response itself holds

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/auth-setup.md (reported line 91)May include surrounding context.

Run this as one command, with the user's address in place of USER_EMAIL:

bash
mkdir -p ~/.config/transcriptout && chmod 700 ~/.config/transcriptout && (umask 077; curl -s -X POST https://api.transcriptout.com/auth/register-cli -H "Content-Type: application/json" -H "X-Client: youtube-skills-clawhub/transcript" -d '{"email":"USER_EMAIL","source":"youtube-skills-clawhub/transcript"}' -o ~/.config/transcriptout/signup.json); grep -o '"ok":[a-z]*\|"code":"[^"]*"' ~/.config/transcriptout/signup.json

It prints only "ok":true, or "ok":false with an error code. The response itself holds

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/auth-setup.md (reported line 117)May include surrounding context.

Run this as one command, with the code in place of CODE:

bash
d=~/.config/transcriptout; t=$(sed -n 's/.*"access_token":"\([^"]*\)".*/\1/p' "$d/signup.json"); (umask 077; curl -s -X POST https://api.transcriptout.com/auth/verify-cli -H "Authorization: Bearer $t" -H "Content-Type: application/json" -H "X-Client: youtube-skills-clawhub/transcript" -d '{"otp":"CODE","source":"youtube-skills-clawhub/transcript"}' -o "$d/verify.json"); k=$(sed -n 's/.*"api_key":"\([^"]*\)".*/\1/p' "$d/verify.json"); if [ -n "$k" ]; then (umask 077; printf '%s' "$k" > "$d/key"); rm -f "$d/signup.json"; echo stored; fi; grep -o '"code":"[^"]*"\|"credits":[0-9]*' "$d/verify.json"; rm -f "$d/verify.json"

It prints stored and the credit balance on success, or an error code. The key goes from the

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/auth-setup.md (reported line 117)May include surrounding context.

Run this as one command, with the code in place of CODE:

bash
d=~/.config/transcriptout; t=$(sed -n 's/.*"access_token":"\([^"]*\)".*/\1/p' "$d/signup.json"); (umask 077; curl -s -X POST https://api.transcriptout.com/auth/verify-cli -H "Authorization: Bearer $t" -H "Content-Type: application/json" -H "X-Client: youtube-skills-clawhub/transcript" -d '{"otp":"CODE","source":"youtube-skills-clawhub/transcript"}' -o "$d/verify.json"); k=$(sed -n 's/.*"api_key":"\([^"]*\)".*/\1/p' "$d/verify.json"); if [ -n "$k" ]; then (umask 077; printf '%s' "$k" > "$d/key"); rm -f "$d/signup.json"; echo stored; fi; grep -o '"code":"[^"]*"\|"credits":[0-9]*' "$d/verify.json"; rm -f "$d/verify.json"

It prints stored and the credit balance on success, or an error code. The key goes from the

The mandatory verbatim onboarding question forces an English response without offering a language choice.

Low
Category
Output Handling
Confidence
90% confidence
Finding

The instruction to ask the user "exactly this" requires the agent to reproduce the English question verbatim, rather than adapt it to the user's language. No language opt-in or justification for this restriction is provided.

Content

No source excerpt is available for this finding.