T09 · Insecure Skill Coding Practices
- Location
scripts/submit-logbook.js:72- Finding
Credentials May Be Disclosed to an Attacker-Controlled Redirect Target
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
The skill is mostly purpose-aligned, but it uses stored MIS credentials to submit live account data and has a real credential-safety flaw that users should review before installing.
Install only if you are comfortable with an agent using stored MIS credentials to submit a live logbook entry. Before use, the script should be hardened to validate the exact login.pens.ac.id HTTPS origin and should require an explicit runtime confirmation before reading credentials or submitting.
scripts/submit-logbook.js:72Credentials May Be Disclosed to an Attacker-Controlled Redirect Target
The skill clearly contemplates use of local secrets for login and browser automation against a live external site, yet it declares no explicit tool scope or permission boundary. That increases the chance an agent can access environment secrets and network capabilities implicitly, making unintended credential use or live submissions more likely.
The activation text is broad enough to trigger during general inspection, preparation, or logbook-related work, not just intentional submission. In a skill that can log in with local secrets and modify data on a live system, over-broad triggering raises the risk of accidental execution and unintended submissions.
The description omits a clear warning that the skill uses local secrets and performs automated actions on a live MIS system. Without that disclosure, users may invoke the skill without realizing it can authenticate and write records, increasing the chance of unintended account use or unauthorized submissions.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
2. Send the draft to the user first.
3. Wait for explicit confirmation.
4. Only then submit the saved draft to MIS.
5. If there is no confirmation yet, do not submit.
## Proven navigation path
The instruction requires the generated summary to be in Indonesian, but the file does not indicate that the user chose this language or that the skill is explicitly limited to Indonesian-only operation. This is a natural-language locale constraint and can violate policy when imposed without opt-in or clear justification.
The script reads credentials from a fixed user-specific secrets file and then uses them to authenticate and perform actions on the user's behalf. In this skill context, that capability is necessary for submission, but it is still security-sensitive because any invocation of the script can trigger privileged use of stored credentials without an explicit per-run consent boundary or narrower secret scoping.
The script accesses local credentials and performs an automated state-changing submission with no user-facing warning, approval prompt, or disclosure at execution time. In an approval-first workflow, silently combining secret access and submission increases the risk of unintended account actions and makes abuse harder for a user to detect before it happens.
The instruction "Write Kegiatan/Materi in natural Indonesian" forces a specific language choice. The file does not state that the user can opt into another language, nor does it document a clear locale-policy justification within the skill text.
The date formatter explicitly uses the 'en-GB' locale, which forces a specific language/locale behavior in the skill. There is no user opt-in, configurability, or comment explaining why this locale is required for interoperability with the target system.