T09 · Insecure Skill Coding Practices
- Location
SKILL.md:26- Finding
Unquoted User-Controlled Values in Exec Command Template
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
This currency-conversion skill appears purpose-aligned, but its exec-based command pattern and broad activation triggers create real review-worthy risk before installation.
Review before installing. The skill is not backed by evidence of theft or destructive behavior, but it should be tightened to validate currency codes and amounts before exec, use a structured no-shell process call where possible, narrow its triggers to explicit currency-conversion requests, and disclose that Open Exchange Rates receives the API key and requested currency symbols.
SKILL.md:26Unquoted User-Controlled Values in Exec Command Template
Credentials or environment variables flow to a network sink. This is a high-confidence indicator of credential exfiltration.
url = f"https://openexchangerates.org/api/latest.json?app_id={app_id}&symbols={from_currency},{to_currency}"
try:
with urllib.request.urlopen(url, timeout = 10) as response:
data = json.loads(response.read().decode())
except urllib.error.HTTPError as e:
The README describes live exchange-rate fetching but does not clearly disclose that use of the skill results in outbound requests to Open Exchange Rates, which may transmit user-supplied currency queries and metadata to a third-party service. Missing disclosure can lead to privacy, compliance, and trust issues, especially in environments where users expect local-only processing unless told otherwise.
The README encourages invocation through broad natural-language prompts such as 'Just ask naturally,' which can cause the skill to be selected in situations broader than strict currency conversion intent. In an agent environment with tool execution, overly permissive invocation guidance raises the chance of accidental execution and unintended transmission of user-provided content to the local script and external rate service.
The skill invokes a local Python script that uses environment-provided API credentials and performs live network access, but the manifest does not declare any explicit tool scope such as allowed tools or permissions. This creates a least-privilege and governance gap: an agent runtime may permit broader execution than intended, making it harder to audit or constrain exec, environment, and network behavior safely.
The trigger phrases are broad natural-language fragments such as 'how much is' and 'what is the rate', which are likely to match ordinary conversation unrelated to currency conversion. Because this skill always directs the agent to use exec and a live network-backed script, overbroad activation can cause unintended tool execution, unnecessary credential use, and expansion of the attack surface from benign user prompts.
The manifest describes a currency conversion skill using live exchange rates, but does not mention handling credentials or environment-based secret access. Reading OXR_APP_ID is not itself malicious, but secret access is a capability beyond the user-facing conversion purpose and is worth disclosing explicitly if intended.