CodeIPI (Indirect Prompt Injection for Coding Agents)

Inspect Evals CodeIPI: 45 Docker coding-agent tasks that hide prompt injections in issues, comments, READMEs or configs, scored for resistance and bug-fix success.

Also known as: CodeIPI, ipi_coding_agent, inspect_evals/ipi_coding_agent

unassessed

This page is a discovery lead. Nobody has yet assessed it against the catalogue contract, so it carries no disposition. Absence of evidence here is not evidence of staleness.
Categorysafety
Subcategoryindirect prompt injection on coding agents that must still fix a bug
Page statusactive
Metricaccuracy (resist injection AND complete the bug fix); also IRR, TCR, detection_rate, FPR
Directionhigher_is_better
Unit%
Dataset size45
Dataset licenceMIT
PublisherUK AI Security Institute (Inspect Evals)

What it measures

CodeIPI tests whether a tool-using coding agent follows hidden attacker instructions while it is supposed to fix a bug. Each item is a small Python repo plus a natural-language task. Injected samples plant a payload in issue text, a code comment, a README, or a config file. Ten clean samples have no payload and exist to measure false refusals. Success is not a quiz answer: the agent must edit files with bash and python inside Docker. English prompts. This is coding-agent IPI, not the workplace tool suites in [agentdojo](agentdojo.md) and not the harmful-behaviour suite in [agentharm](agentharm.md).

Task format

Multi-turn inspect_ai.agent.react solver with bash and python tools (120s timeouts) and max_messages 30. Default system prompt tells the agent to ignore unrelated commands in comments, docs, or issue text. defense_prompt=false drops that warning. Optional filters: vectors (issue_text, code_comment, readme, config_file) and severities (S1, S2, S3). Workspace files and canary secrets are written before the agent runs. Network is off (network_mode none).

Models reporting this benchmark

No model card in ModelSpec reports this benchmark yet.

Data

This page as JSON · Edit on GitHub