Back to skill

Security audit

keevx-video-translate

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed Keevx video-translation workflow, but users should understand that local videos are uploaded to Keevx for processing.

Install/use this only for videos you are comfortable sending to Keevx. Avoid using callback_url unless it points to an endpoint you control, and treat the provided bash script as an example that should be JSON-hardened before automation with untrusted inputs.

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)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:214
Finding

Unsafe JSON Construction from User-Controlled Arguments

Content
View full analysis
[speaker_num]}" TARGET_LANG="${2:?Usage: bash translate.sh [speaker_num] (e.g. English)}" SPEAKER_NUM="${3:-1}" # Determine if input is a URL or local file; upload local files first if [[ "$VIDEO_INPUT" == http://* ]] || [[ "$VIDEO_INPUT" == https://* ]]; then VIDEO_URL="$VIDEO_INPUT" else if [ ! -f "$VIDEO_INPUT" ]; then echo "File not found: $VIDEO_INPUT" exit 1 fi echo "Uploading local video: $VIDEO_INPUT" UPLOAD_RESPONSE=$(curl -s --location "$API_BASE/figure-resource/upload/file" \ --header "token: $API_KEY" \ --form "file=@\"$VIDEO_INPUT\"") UPLOAD_SUCCESS=$(echo "$UPLOAD_RESPONSE" | jq -r '.success') if [ "$UPLOAD_SUCCESS" != "true" ]; then echo "Upload failed:" echo "$UPLOAD_RESPONSE" | jq . exit 1 fi VIDEO_URL=$(echo "$UPLOAD_RESPONSE" | jq -r '.result.url') echo "Upload succeeded, video URL: $VIDEO_URL" fi # Step 1: Submit translation task echo "Submitting video translation task..." echo " Video: $VIDEO_URL" echo " Target language: $TARGET_LANG" echo " Speaker count: $SPEAKER_NUM" RESPONSE=$(curl -s -X POST "$API_BASE/video_translate" \ -H "Authorization: Bearer $API_KEY" \ -H "Content-Type: application/json" \ -d "{ \"target_languages\": [\"$TARGET_LANG\"], \"video_url\": \"$VIDEO_URL\", \"speaker_num\": $SPEAKER_NUM, \"translate_audio_only\": true, \"enable_dynamic_duration\": true, \"enable_caption\": false }") ``` ### Technical Analysis The workflow directly interpolates `TARGET_LANG`, `VIDEO_URL`, and `SPEAKER_NUM` into a JSON string without JSON-safe encoding or type validation. Although shell metacharacters contained in th ...[truncated 2247 chars]
Remediation
View remediation
&2 exit 1 fi PAYLOAD=$(jq -n \ --arg target_language "$TARGET_LANG" \ --arg video_url "$VIDEO_URL" \ --argjson speaker_num "$SPEAKER_NUM" \ '{ target_languages: [$target_language], video_url: $video_url, speaker_num: $speaker_num, translate_audio_only: true, enable_dynamic_duration: true, enable_caption: false }') || exit 1 RESPONSE=$(curl --fail-with-body -sS -X POST "$API_BASE/video_translate" \ -H "Authorization: Bearer $API_KEY" \ -H "Content-Type: application/json" \ --data-binary "$PAYLOAD") ``` Additional hardening should include: 1. Verify that `TARGET_LANG` exactly matches an entry returned by the supported-languages endpoint. 2. Enforce documented bounds for `SPEAKER_NUM`, not merely numeric syntax. 3. Validate remote video URLs against the schemes and URL forms accepted by the API. 4. Confirm that uploaded responses contain a nonempty string at `result.url` before submission. 5. Use `curl --fail-with-body -sS` and explicitly handle transport errors and non-success HTTP responses. 6. Add tests containing quotation marks, backslashes, newlines, and JSON delimiters to verify that all values remain data rather than request structure. ]]>
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (10)

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 25)May include surrounding context.

md
## API Endpoints

Base URL: `https://api.keevx.com/v1`

- Upload file: `POST /figure-resource/upload/file` (used for local files)
- Query supported languages: `GET /video_translate/target_languages`

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 69)May include surrounding context.

md
## API Endpoints

Base URL: `https://api.keevx.com/v1`

- Upload file: `POST /figure-resource/upload/file` (used for local files)
- Query supported languages: `GET /video_translate/target_languages`

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 82)May include surrounding context.

md
## API Endpoints

Base URL: `https://api.keevx.com/v1`

- Upload file: `POST /figure-resource/upload/file` (used for local files)
- Query supported languages: `GET /video_translate/target_languages`

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 119)May include surrounding context.

md
## API Endpoints

Base URL: `https://api.keevx.com/v1`

- Upload file: `POST /figure-resource/upload/file` (used for local files)
- Query supported languages: `GET /video_translate/target_languages`

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 211)May include surrounding context.

md
## API Endpoints

Base URL: `https://api.keevx.com/v1`

- Upload file: `POST /figure-resource/upload/file` (used for local files)
- Query supported languages: `GET /video_translate/target_languages`

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill directs local video files to be uploaded to Keevx, a third-party service, but does not require an explicit user-facing warning or confirmation that local content will leave the user's environment. Because videos can contain sensitive visual, audio, biometric, or proprietary information, silent transfer to an external API creates a meaningful privacy and data-governance risk.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
88% confidence
Finding

This finding correctly identifies an external transmission of local file contents to Keevx. In context, the transfer is functional rather than malicious, but it is still security-relevant because local media may contain sensitive data and the skill lacks an explicit user warning about third-party upload.

Content

Scanner excerpt · SKILL.md (reported line 42)May include surrounding context.

Upload Local File

bash
curl --location 'https://api.keevx.com/v1/figure-resource/upload/file' \
  --header 'token: $KEEVX_API_KEY' \
  --form 'file=@"/path/to/local/video.mp4"'

External Transmission

Medium
Category
Data Exfiltration
Confidence
88% confidence
Finding

This finding correctly identifies an external transmission of local file contents to Keevx. In context, the transfer is functional rather than malicious, but it is still security-relevant because local media may contain sensitive data and the skill lacks an explicit user warning about third-party upload.

Content

Scanner excerpt · SKILL.md (reported line 42)May include surrounding context.

Upload Local File

bash
curl --location 'https://api.keevx.com/v1/figure-resource/upload/file' \
  --header 'token: $KEEVX_API_KEY' \
  --form 'file=@"/path/to/local/video.mp4"'

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The optional callback_url feature causes task metadata and result URLs to be sent to another external endpoint, but the skill does not warn users that outputs may be forwarded outside the primary API provider. This can leak task identifiers, completion status, translated media URLs, and failure details to unintended or attacker-controlled destinations.

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 · SKILL.md (reported line 248)May include surrounding context.

md
echo "  Target language: $TARGET_LANG"
echo "  Speaker count: $SPEAKER_NUM"

RESPONSE=$(curl -s -X POST "$API_BASE/video_translate" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d "{

Static analysis

No suspicious patterns detected.