Back to skill

Security audit

Artifact Analysis

Security checks for vulnerabilities and agentic risk

Overview

This local document-analysis skill is mostly coherent, but it needs Review because it advises users to paste sensitive credential-file contents directly into the agent if skipped files are needed.

Review before installing. Use explicit paths and output_dir values, avoid no-parameter auto-discovery on private projects, and do not paste private keys, tokens, .env contents, cloud credentials, session cookies, or complete credential files into the agent; provide only sanitized, redacted excerpts when needed.

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
references/skip-patterns.md:49
Finding

Unsafe Guidance Encourages Plaintext Disclosure of Sensitive Credentials

Content
View full analysis

Vulnerability Details

File Location: references/skip-patterns.md:49
Vulnerability Type: Plaintext sensitive-data exposure through unsafe user guidance
Risk Level: Medium

Vulnerable Code Snippet

markdown
- **Sensitive list is non-overridable.** `.env`, `*.pem`, `*.key`, and credential files are always skipped even when explicitly passed. If a caller needs those contents, they paste them in directly rather than routing through this skill.

Technical Analysis

The skill correctly prevents filesystem scanning of sensitive files such as environment files, private keys, and credential files. However, its documented fallback tells callers to paste those contents directly into the conversation.

This guidance circumvents the denylist's security purpose. Pasted secrets can enter model context, conversation logs, telemetry, or downstream processing. The issue does not constitute an automatic SSH-key operation, credential read, or exfiltration mechanism; it is an insecure usage instruction that may cause users to disclose sensitive material manually.

An attacker or untrusted document could exploit this behavior socially by claiming that omitted credential contents are necessary for analysis and directing the user to follow the documented fallback.

Attack Path

  1. A caller requests analysis involving a sensitive file such as .env, an SSH private key, or a credential file.
  2. The skill correctly excludes that file under its non-overridable denylist.
  3. The caller follows the documented instruction to paste the file contents directly.
  4. Plaintext tokens, passwords, cloud credentials, certificates, or private keys enter the agent's conversational context.
  5. Those secrets may subsequently be retained in logs, included in downstream processing, or exposed to any system or person with access to the conversation.

Impact Assessment

Successful exploitation requires the caller to follow the unsafe instruction; the skill does not inde ...[truncated 549 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove the instruction to paste sensitive files or credentials into the conversation.
  2. Replace it with explicit guidance to provide only a sanitized copy or narrowly scoped redacted excerpt.
  3. Warn callers never to submit private keys, passwords, access tokens, session cookies, recovery codes, or complete credential files.
  4. Require secret values to be replaced with conspicuous placeholders such as REDACTED_API_TOKEN.
  5. Preserve only non-secret structure when needed for analysis, for example variable names without values.
  6. Add a pre-processing rule that rejects or warns on likely private-key blocks, token formats, and unredacted credential assignments in user-provided text.
  7. Ensure sensitive text is not forwarded to subagents, findings files, reports, logs, or persistent memory.
  8. Document a secure alternative workflow, such as creating a temporary sanitized copy outside the sensitive path and scanning only that copy after user verification.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (6)

Context Leakage

High
Category
Data Exfiltration
Confidence
90% confidence
Finding

The skill is explicitly designed to scan local documents and project knowledge, auto-discover common documentation locations, and synthesize their contents into reports on disk. In an agent setting, broad context extraction can expose sensitive internal information from local files beyond what a user intended to share, especially when invocation is triggered by broad phrases or by other skills programmatically.

Content

Scanner excerpt · SKILL.md (reported line 3)May include surrounding context.

md
---
name: artifact-analysis
description: "Use when the user wants a cited, structured read of local documents and project knowledge. Triggers on: \"analyze these docs\", \"scan my project for context\", \"read the docs folder\", \"summarize what's in .beagle/concepts/\", \"extract context from docs/\", \"what's in this folder\", \"go read everything in X and tell me what's there\". Also invoked programmatically by other beagle skills (prfaq-beagle Ignition, brainstorm-beagle reference points, strategy-interview context grounding) via the companion contract. Does NOT trigger on codebase lookups (\"find this function\", \"search the repo\"), web research (use web-research), LLM-as-judge evaluation (use llm-judge), or document editing (use humanize-beagle). Produces a written scan plan, parallel-subagent findings, and a cited synthesis report on disk — never inline prose, never unsourced claims."
---

# Artifact Analysis

Credential Access

High
Category
Privilege Escalation
Confidence
80% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/skip-patterns.md (reported line 15)May include surrounding context.

md
- `*.pem`, `*.key`, `*.p12`, `*.pfx`, `*.crt`, `*.cer`
- `.aws/`, `.ssh/`, `.gnupg/`
- `credentials.*`, `secrets.*`
- `*.htpasswd`, `*.netrc`

## Binary / media

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

This markdown skill clearly states that deliverables are written on disk, and later describes overwriting behavior plus archival of prior contents under the same output directory. However, it does not present this as a user-facing warning or caution despite affecting the filesystem and prior run artifacts.

Content

No source excerpt is available for this finding.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · SKILL.md (reported line 23)May include surrounding context.

md
- Codebase lookups ("where is this function defined", "grep for this symbol"). Search the codebase instead.
- Web research. Use [web-research](../web-research/SKILL.md).
- Comparative evaluation of two implementations or source credibility adjudication. Use [llm-judge](../llm-judge/SKILL.md).
- Rewriting or editing the scanned documents. Use humanize-beagle ([../../../beagle-docs/skills/humanize-beagle/SKILL.md](../../../beagle-docs/skills/humanize-beagle/SKILL.md)) or edit the files directly.
- PDF / image OCR / format conversion. First version reads plain text and markdown only.
- Paywalled or authentication-gated remote sources. This is a local-filesystem primitive.
- Coaching, challenge, or reshaping of the caller's question. That belongs to the caller.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The markdown explicitly states 'Multi-language analysis. English-only today.' That is a language policy constraint expressed in natural language, but the skill does not offer a language choice or frame the restriction as user opt-in.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

This markdown file includes a standalone direct-trigger example where the user says "analyze these docs" and "read everything in docs/ and tell me what's there," and even labels it as "Triggered by user chatter." Those phrases are common conversational requests rather than narrowly scoped invocation commands, and the contract does not provide exclusion conditions or constraints to prevent accidental activation.

Content

No source excerpt is available for this finding.