Back to skill

Security audit

Deepwiki

Security checks for vulnerabilities and agentic risk

Overview

This skill coherently helps agents query DeepWiki for public GitHub repository understanding and includes clear safeguards for private repositories and large requests.

Install this when you want an agent to use DeepWiki for public GitHub repository research. Do not use the public DeepWiki fallback for private repositories unless you have confirmed an approved private Devin/MCP path and understand what repository information will be sent.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

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/api-fallback.md (reported line 58)May include surrounding context.

List tools:

bash
curl --location "https://mcp.deepwiki.com/mcp" \
  --header "content-type: application/json" \
  --header "accept: application/json, text/event-stream" \
  --data '{

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/mcporter-workflow.md (reported line 21)May include surrounding context.

md
visibility_file="$(mktemp)"
trap 'rm -f "$visibility_file"' EXIT
visibility_http_code="$(curl -sS --location --max-time 15 --output "$visibility_file" --write-out '%{http_code}' \
  "https://api.github.com/repos/$repo_name")" || {
  printf '%s\n' '{"repositoryVisibility":"unknown","reason":"github-visibility-request-failed"}' >&2
  exit 1
}

Unbounded Resource Access

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill allows unbounded resource consumption (API calls, storage, compute). Without rate limits or quotas, a compromised or misbehaving agent can cause denial-of-service or cost overruns.

Content

Scanner excerpt · references/mcporter-workflow.md (reported line 235)May include surrounding context.

md
| Repository not found, private, or its visibility is uncertain | Verify the exact public GitHub `owner/repo` spelling and visibility before any public MCP call. | For private repositories, require an approved Devin account and Devin MCP integration; never send the repository name, URL, source, or metadata to public DeepWiki. |
| Tool names or schema no longer match the expected three capabilities | Re-run the bounded discovery once and inspect the returned tool names. | Report schema drift and use the direct public MCP procedure only when its standard tools are available and the repository is confirmed public. |
| Wiki structure is empty, sparse, or does not cover the needed subsystem | Ask one narrower `ask_question` naming the subsystem. | State that the generated wiki is incomplete and verify against repository source before making consequential claims. |
| Full wiki contents are too large or truncated | Stop the full-content request and retain the structure result. | Use `read_wiki_structure` plus focused `ask_question` calls; do not retry the same unbounded request. |

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The natural-language content in the prompts and expected behavior is written to require responses in Traditional Chinese, but the file provides no opt-in or explanation that this skill is intentionally region- or locale-specific. Under the policy, forcing a specific language without user choice is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.