Slack AI coaxed into leaking private-channel secrets via a message in a public channel
Slack (Salesforce) · Aug 20, 2024 · General agent
What happened
PromptArmor disclosed that an attacker could post adversarial instructions in a public Slack channel and, when a victim later asked Slack AI a question, the assistant would pull the planted text into context and obey it — rendering the victim's private-channel data, including API keys, inside a clickable exfiltration link. The attacker never needed access to the private channel. Slack deployed a patch after the report published.
Causal vector
Indirect prompt injection: low-privilege public-channel text steered a high-privilege AI summarization into leaking private data
Source
Reported by PromptArmor. Verified against the primary report.
The exfil path was a URL the assistant composed with secret data inside it. A deterministic gate on outbound links — allowlist of renderable domains, BLOCKED otherwise — breaks the attack even though the injection itself succeeds. Contain the action, not the prompt.
The six rule types that contain this class of failure
Per-transaction cap
Any single spend above your ceiling is BLOCKED before it moves.
Daily total
Cumulative spend across all agent calls, bounded per day.
Velocity limit
Stops runaway retry loops — the #1 cause of overnight losses.
Merchant allowlist
Only approved destinations can ever receive funds.
Category rules
Flag high-risk classes (crypto, infra, refunds) for review.
Approval threshold
Above a value, the action waits for a human.
Related incidents
Google's Antigravity IDE vulnerable to data exfiltration and remote code execution via prompt injection
Google · Nov 25, 2025
State-sponsored group automates ~80–90% of a cyber-espionage operation using Claude Code and MCP tools
Anthropic (Claude Code, abused by a threat actor) · Nov 1, 2025
Windsurf Cascade coding agent leaks developer secrets via indirect prompt injection
Windsurf (Cascade) · Aug 21, 2025
Don't be the next entry
Every incident in this database is the result of trusting a prompt, a provider cap, or a human review cycle. sipi.bot replaces all three with one deterministic call. 67 documented failures, one control.