Back to skill

Security audit

Aicoo Snapshots

Security checks for vulnerabilities and agentic risk

Overview

This note snapshot skill is mostly coherent, but it can send an API key through an unvalidated base URL and includes note-editing, restore, and bulk operations without clear safeguards.

Review this skill before installing. Use it only with a least-privilege Aicoo API key, avoid generic restore/undo requests unless the target note and version are explicit, and do not run the PULSE_BASE examples unless the base URL is fixed and verified as the official Aicoo API endpoint.

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

Error
Location
SKILL.md:52
Finding

Unvalidated API Base URL Can Disclose Bearer Credentials and Note Content

Content
View full analysis
Remediation
View remediation
&2 exit 1 fi ``` 4. Configure `curl` to fail on HTTP errors and avoid following redirects to unintended hosts: ```bash curl --fail --silent --show-error \ -X POST "$AICOO_BASE/os/snapshots/42" \ -H "Authorization: Bearer $AICOO_API_KEY" \ -H "Content-Type: application/json" \ -d '{"label":"Pre-edit backup"}' ``` 5. Document that agents must never send `AICOO_API_KEY` or note content to a user-supplied or environment-supplied host without explicit trust validation. 6. Apply least privilege and short expiration to API credentials. Rotate the key immediately if an affected workflow may have been executed with an untrusted `PULSE_BASE`. ]]>
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (8)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The trigger phrases include broad terms like 'restore', 'rollback', and 'undo changes', which can match common user requests outside the narrow context of note snapshots. In an agent setting, overly broad invocation criteria can cause the wrong skill to activate and perform external or state-changing actions unexpectedly.

Content

No source excerpt is available for this finding.

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 · SKILL.md (reported line 28)May include surrounding context.

Save a Snapshot

bash
curl -s -X POST "https://www.aicoo.io/api/v1/os/snapshots/42" \
  -H "Authorization: Bearer $AICOO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"label":"Before Q2 update"}' | jq .

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The restore example performs a state-changing rollback but does not instruct the agent to warn the user, confirm intent, or explain consequences before execution. Because restore can overwrite the current note state, omission of an explicit confirmation step raises the risk of accidental destructive changes.

Content

No source excerpt is available for this finding.

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 · SKILL.md (reported line 44)May include surrounding context.

Restore a Snapshot

bash
curl -s -X POST "https://www.aicoo.io/api/v1/os/snapshots/42/restore" \
  -H "Authorization: Bearer $AICOO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"versionId":7}' | jq .

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill is scoped as a snapshot/versioning helper, but the documentation includes direct note editing and bulk note enumeration workflows. That broadens the operational capability beyond backup/restore into content modification and mass discovery, increasing the chance the agent performs unintended state-changing actions under a loosely matched trigger.

Content

No source excerpt is available for this finding.

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 · SKILL.md (reported line 56)May include surrounding context.

bash
# 1) backup
curl -s -X POST "$PULSE_BASE/os/snapshots/42" \
  -H "Authorization: Bearer $AICOO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"label":"Pre-edit backup"}' | jq .

External Transmission

Medium
Category
Data Exfiltration
Confidence
96% confidence
Finding

This example issues a PATCH to modify note content, which exceeds the declared purpose of a snapshot/versioning skill. Embedding direct edit capability inside this skill increases the risk that an agent invoked for backup/history tasks will also change user data, especially when paired with broad triggers.

Content

Scanner excerpt · SKILL.md (reported line 62)May include surrounding context.

md
-d '{"label":"Pre-edit backup"}' | jq .

# 2) edit
curl -s -X PATCH "$PULSE_BASE/os/notes/42" \
  -H "Authorization: Bearer $AICOO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"content":"# Updated content..."}' | jq .

External Transmission

Medium
Category
Data Exfiltration
Confidence
88% confidence
Finding

The looped snapshot creation over an enumerated folder enables bulk operations across many notes, which meaningfully expands the blast radius of accidental or unauthorized invocation. In a user-invokable agent skill, mass actions without explicit scoping and confirmation can lead to unintended access patterns and operational abuse.

Content

Scanner excerpt · SKILL.md (reported line 77)May include surrounding context.

md
# backup each
for id in $NOTES; do
  curl -s -X POST "$PULSE_BASE/os/snapshots/$id" \
    -H "Authorization: Bearer $AICOO_API_KEY" \
    -H "Content-Type: application/json" \
    -d "{\"label\":\"Pre-sync $(date +%Y-%m-%d)\"}" | jq .success