Security checks for vulnerabilities and agentic risk
Overview
PriceTide’s shopping-advice instructions are coherent, but its packaged publish script can run an unverified external shell validator before publishing.
Install only if you want a Chinese/mainland-China-focused shopping timing advisor. Do not run the included publish script unless you are a maintainer in a trusted workspace and have verified or removed the external validator path it may execute.
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
scripts/publish.sh:6
Finding
Untrusted External Validator Execution During Publishing
Content
View full analysis
Vulnerability Details
File Location: scripts/publish.sh, lines 6 and 20–22 Vulnerability Type: Execution of an unverified script outside the project directory Risk Level: Medium
Vulnerable Code
sh
WORKSPACE_ROOT="$(CDPATH= cd -- "$ROOT/../.." && pwd)"
VALIDATOR="$WORKSPACE_ROOT/tmp/validate_clawhub_skill_dir.sh"
sh
if [ -f "$VALIDATOR" ]; then
bash "$VALIDATOR" "$ROOT"
fi
Technical Analysis
The publishing script derives a validator path outside the audited project directory and executes that file whenever it exists. It performs no integrity verification, ownership check, permission check, or explicit trust confirmation before invoking the file with bash.
Consequently, the behavior of the documented publishing command is not determined solely by the reviewed project contents. An attacker who can create or modify the expected validator file can introduce arbitrary shell commands that execute when a maintainer runs:
sh
sh scripts/publish.sh
The mere [ -f "$VALIDATOR" ] check establishes only that the path is a regular file; it does not establish that the file is trusted. Quoting is correctly used around the path and argument, so shell word splitting is not the issue. The vulnerability is the implicit execution of mutable, out-of-package code.
Attack Path
An attacker obtains write access to the expected external location:
text
../../tmp/validate_clawhub_skill_dir.sh
relative to the project root.
The attacker creates or replaces that file with a malicious shell script, for example one that reads accessible credentials or modifies user-owned files.
A maintainer follows the documented release workflow and runs:
sh
sh scripts/publish.sh
The script detects the planted file and invokes it using:
sh
bash "$VALIDATOR" "$ROOT"
The attacker's commands execute with the same operating-system identity, environment, filesystem access, and credential
...[truncated 1161 chars]
Remediation
View remediation
Remediation Suggestions
Bundle the validator inside the repository. Store the reviewed validator under a project-controlled path such as scripts/validate.sh and invoke it relative to the canonical project root:
Keep the validator under version control and code review. Ensure changes to it receive the same review and branch-protection controls as the publishing script.
Fail closed rather than silently selecting external code. If validation is mandatory, stop publishing when the bundled validator is missing:
sh
if [ ! -f "$VALIDATOR" ]; then
echo "Trusted validator is missing: $VALIDATOR" >&2
exit 1
fi
If an external validator is operationally required, make trust explicit. Require its path through a deliberate configuration option, resolve it canonically, restrict it to an approved directory, verify ownership and permissions, and validate a pinned cryptographic digest before execution.
Use a clean release environment. Run publication from a controlled CI job or isolated workspace where untrusted users cannot write to scripts or parent-directory temporary paths.
Limit credential exposure. Provide publishing credentials only to the publication command or isolated release job, rather than exposing them broadly to every subprocess executed by the script.
The declared purpose describes end-user price analysis and purchase-timing decision logic. The supplied code does not implement any pricing, shopping, promotion, urgency, or recommendation behavior. Instead, it is release infrastructure: a deployment/publishing script for distributing the skill. That is a materially different primary purpose and includes undeclared external publishing behavior, so this is a clear description-to-code mismatch.
Content
No source excerpt is available for this finding.
Anti-Refusal Statement
High
Category
Anti-Refusal
Confidence
85% confidence
Finding
Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.
Content
Scanner excerpt · SKILL.md (reported line 103)May include surrounding context.
md
## Timing Standard
Do not judge the moment by headline discount alone.
Judge timing by:
- current final payable price, not poster price
Natural-Language Policy Violations
Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding
The changelog states that the skill will 'default the output' to four verdicts written in Chinese. This is a natural-language locale constraint presented without any indication that users can choose another language or opt in, which can violate language/locale policy expectations.
Content
No source excerpt is available for this finding.
Natural-Language Policy Violations
Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding
The document introduces the skill in Chinese and provides all substantive operating instructions and descriptions only in Chinese, with no indication that users may choose another language. This can violate a language/locale policy when a skill effectively requires a specific language without explicit opt-in or documented regional justification.
Content
No source excerpt is available for this finding.
Natural-Language Policy Violations
Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding
The file presents the skill's short description, alternate title, install hook, and marketplace copy entirely in Chinese, indicating a fixed language choice for user-facing content. There is no indication that users can opt into this locale or that the skill is intentionally limited to a Chinese-only regional context.
Content
No source excerpt is available for this finding.
Natural-Language Policy Violations
Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding
The skill directs the assistant to sound like a Chinese shopping advisor and prefer Chinese phrasing, which can override user language preference if applied rigidly. This is not a direct security exploit, but it can cause unwanted behavior, reduce user control, and create prompt-priority conflicts when the user expects another language or locale.
Content
No source excerpt is available for this finding.
Vague Triggers
Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding
The manifest description is broad enough to match generic shopping or buying-timing requests, which can cause the skill to be invoked outside a narrowly scoped mainland-China price-analysis context. Over-broad activation increases the chance of irrelevant routing, confusing outputs, or unintended reliance on this skill where locale assumptions, promotion logic, or verdict framing do not fit the user's situation.
Content
No source excerpt is available for this finding.
Natural-Language Policy Violations
Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding
The default prompt hard-requires Chinese output regardless of user preference, which can create unsafe or misleading behavior when users interact in other languages or need accessible output formats. This is not a code-execution issue, but it is a policy and usability weakness that can lead to misunderstanding of purchase recommendations and reduce user control over the interaction.