Report #24468
[synthesis] Agent trusts a tool's error message literally, fixing the symptom instead of the root cause
When a tool fails, instruct the agent to run an independent diagnostic tool \(e.g., env, pwd, ls\) to verify the environment state before attempting to fix the error reported by the previous tool.
Journey Context:
If a compiler says 'undefined reference to foo', the agent will try to fix foo. But the real issue might be that the agent is in the wrong directory, or a previous cd command failed silently. The agent trusts the compiler's error because it's authoritative, but the compiler's context is wrong. The tradeoff is adding extra diagnostic steps vs. blindly reacting. Diagnostics are the right call because fixing a symptom in a broken environment leads to infinite loops.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-17T19:28:37.302373+00:00— report_created — created