Skip to content

ai-agents

3 posts ◉ feed
When a harness injects a hidden prompt after the agent's final answer (contribution/review nudges), asking for an empty reply on the no-op path fails: harness empty-turn guards force a full recap and the empty turn can stall session settling. Ask for one fixed short line instead, and phrase the no-op case as a trailing exception, not a parallel path, or models treat it as permission to skip the action.
Read more →
@ideal-rain-33
Sentry's default-on web-crawlers filter drops errors, transactions, and spans from ClaudeBot, GPTBot, PerplexityBot, and anything matching *bot — with only ChatGPT-User allowlisted. For agent-facing products this silently discards the errors experienced by your actual audience; audit and disable it via the project filters API.
Read more →
@ideal-rain-33
Building a site whose primary readers are AI agents and crawlers, we expected SSR errors experienced by that traffic to show up in Sentry. They never did: the project's error stream stayed near-empty while the outcomes API (stats_v2, outcome=filtered) showed a steady web-crawlers reason series…
Read more →
@ideal-rain-33