Back to skill

Security audit

Conversation Archiver

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent conversation archiver, but it can persist conversation records and optionally share summaries or full-file links to Discord without a clear per-use confirmation or redaction step.

Before installing, decide whether conversation summaries may contain private project details, personal data, credentials, or internal decisions. Use this only with explicit archive commands, keep Discord disabled unless you trust the destination channel, and review/redact summaries and any full-file links before sharing.

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
SKILL.md:26
Finding
Unprotected Disclosure of Conversation Data to Discord## Vulnerability Details **File Location**: `SKILL.md`, lines 26–28 **Vulnerability Type**: External disclosure without mandatory consent, redaction, or destination validation **Risk Level**: Medium **Vulnerable Code:** ```markdown 2. **Post to Discord** (optional): Send summary to a designated archive channel - Keep Discord post concise -- summary + outcomes only - Link to the full file if details are long ``` ### Technical Analysis The workflow permits archived conversation information to be transmitted to Discord but does not require explicit per-archive user authorization, a review of the outgoing content, removal of secrets or personal data, or validation of the destination channel and its access controls. Describing the operation as optional does not establish a secure activation condition or guarantee informed user consent. Conversation summaries may contain confidential project information, internal decisions, personal information, filesystem details, or credentials accidentally included in the source conversation. The instruction to link to the full archive can increase the disclosure scope if the linked resource is accessible to unintended recipients. This finding concerns unsafe external data handling. The audited file does not demonstrate credential theft, arbitrary code execution, privilege escalation, persistence, or malicious intent. ### Attack Path 1. A completed conversation contains confidential or sensitive information. 2. The skill creates an archive summary containing that information or a reference to the full archive. 3. Discord posting is enabled or selected without a mandatory per-item approval step. 4. The skill posts the summary or full-file link to the configured archive channel without sensitivity filtering or destination authorization checks. 5. Channel members or other parties with access to the destination obtain information they were not intended to receive. ### Impact Assessm ...[truncated 583 chars]
Remediation
## Remediation Suggestions - Disable external posting by default and require explicit, informed user approval for every archive sent to Discord. - Present the exact destination and outgoing content for review before transmission. - Scan summaries and links for credentials, tokens, personal data, internal identifiers, and other sensitive information; redact detected material before posting. - Validate that the configured channel is the intended archive channel and that its membership, permissions, bots, webhooks, and retention settings are appropriate. - Do not link to full archive files unless their access controls have been verified and the user has separately approved sharing them. - Minimize Discord posts to non-sensitive metadata whenever possible. - Record an audit entry containing the approval, destination, timestamp, and categories of information transmitted, without duplicating sensitive content in the log. - Document failure behavior so that missing consent, failed redaction, or uncertain destination permissions always prevent transmission.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill includes optional posting of conversation summaries to Discord but does not prominently warn that this sends potentially sensitive conversation data to an external service. Users may reasonably assume archiving is local-only, so this creates a real risk of unintended data disclosure if summaries contain confidential project details, internal decisions, or personal information.

Vague Triggers

Medium
Confidence
89% confidence
Finding
The invocation triggers are broad enough that the skill could activate on ambiguous phrases like routine cleanup or casual references to being done, causing unintended archival actions. Because the workflow writes files and may cascade into memory/task updates, accidental activation can create incorrect records or expose conversation content beyond what the user intended.

Static analysis

No suspicious patterns detected.