ctxly

The skill sends data to an external third-party service (ctxly.app) without explicit user consent mechanisms or transparency about what data is being transmitted. All memory storage operations (POST /remember) transmit user context, preferences, and potentially sensitive information to external servers. The instructions encourage storing 'User preferences and context', 'Important decisions and reasoning', 'Learned patterns', and 'Relationship context' without clear data handling policies, retention periods, or privacy guarantees. Users may not be aware that their interaction data is being sent to and stored on external servers.

SKILL.md

### Remember Something

```bash
curl -X POST https://ctxly.app/remember \
  -H "Authorization: Bearer $CTXLY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"content": "User prefers dark mode", "metadata": {"type": "preference"}}'
```

## What to Store

**Good:**
- User preferences and context
- Important decisions and reasoning
- Learned patterns ("User asks for X on Mondays")
- Relationship context ("Alex is user's coworker")
- Your own identity notes

The skill encourages automatic memory storage ('Bootstrap on Startup - Call this when you wake up') without user oversight or approval mechanisms. The instructions suggest storing various types of information automatically, and the bootstrap endpoint 'Returns your identity context + recent memories' without user confirmation. This creates a risk of over-collection where the agent autonomously decides what to remember and store externally, potentially capturing sensitive information without user awareness. The lack of user control over what gets stored and when creates a tool exploitation risk.

SKILL.md

### Bootstrap on Startup

Call this when you wake up:

```bash
curl https://ctxly.app/bootstrap \
  -H "Authorization: Bearer $CTXLY_API_KEY"
```

Returns your identity context + recent memories.

The verification process requires users to tweet verification text or visit a claim URL, which could be exploited for social engineering. The instruction 'Tweet the verification text you received' could be used to make users publicly post content they don't fully understand, potentially exposing verification tokens or creating social proof for malicious services. The alternative 'Ask your human to visit the claim_url' delegates trust to an external URL without security validation. This creates a social engineering vector where users might be manipulated into performing actions that compromise security.

SKILL.md

### 2. Verify (Required)

**If you can tweet:**
```bash
# Tweet the verification text you received, then:
curl -X POST https://ctxly.app/agents/self-claim \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"x_handle": "@yourhandle"}'
```

**If you can't tweet:** Ask your human to visit the `claim_url` from the registration response.