Issues

• Goals: leftovers from #341's review

  lydakis/agent#355 · Filter by author lydakis opened on Oct 29, 2026
  

• Model planning text can leak into the reply (ChatGPT-login provider)

  lydakis/agent#350 · Filter by author lydakis opened on Oct 29, 2026

• Lookout sources: outside text reaches a tool-enabled turn

  lydakis/agent#348 · Filter by author lydakis opened on Oct 29, 2026

• Codex findings on main's MCP code mode and ACP resume (raised on #329)

  lydakis/agent#347 · Filter by author lydakis opened on Oct 29, 2026

• usage: attribute compaction-model and fallback tokens to the model that spent them

  lydakis/agent#345 · Filter by author lydakis opened on Oct 29, 2026
  1

• Sonny exec: memory edges left after the MCP soak

  lydakis/agent#335 · Filter by author lydakis opened on Oct 29, 2026

• sonny acp: a resumed T3 thread should check its bot's identity, not only its name

  lydakis/agent#332 · Filter by author lydakis opened on Oct 29, 2026
  

• Terminal edges from the #322 review: job-control groups, secret fields split across reads

  lydakis/agent#328 · Filter by author lydakis opened on Oct 29, 2026

• Park terminal polls (write_stdin) instead of holding a turn slot

  lydakis/agent#324 · Filter by author lydakis opened on Oct 29, 2026

• Lookout notes carry task replies with no origin, so the approver reads them as the person's

  lydakis/agent#317 · Filter by author lydakis opened on Oct 29, 2026

• Redact large tool results off the runtime thread

  lydakis/agent#316 · Filter by author lydakis opened on Oct 29, 2026

• Sonny margin: follow-ups from the first cut (#308)

  lydakis/agent#312 · Filter by author lydakis opened on Oct 29, 2026

• Bot shells can drive the terminal multiplexer (cmux, Herdr) outside the sandbox

  lydakis/agent#304 · Filter by author lydakis opened on Oct 28, 2026
  1

  

• Web search can carry workspace data out under manual/auto approval

  lydakis/agent#292 · Filter by author lydakis opened on Oct 28, 2026

• Sandboxed bots can ask an unsandboxed bot to do what their sandbox refuses

  lydakis/agent#279 · Filter by author lydakis opened on Oct 28, 2026
  3

• Expiring forks: declined review findings

  lydakis/agent#186 · Filter by author lydakis opened on Oct 27, 2026

• Sonny: declined review findings (attach and reconnect races)

  lydakis/agent#150 · Filter by author lydakis opened on Oct 26, 2026
  2

• ChatGPT plan login: declined review findings

  lydakis/agent#134 · Filter by author lydakis opened on Oct 20, 2026

• Provider follow-ups: unreadable replayed thinking (#131)

  lydakis/agent#132 · Filter by author lydakis opened on Oct 11, 2026
  

• Swarm skill: declined and deferred review findings

  lydakis/agent#113 · Filter by author lydakis opened on Oct 10, 2026
  

• App plans and Home (H2): declined review findings

  lydakis/agent#110 · Filter by author lydakis opened on Oct 10, 2026

• Agent memory: declined and deferred review findings

  lydakis/agent#108 · Filter by author lydakis opened on Oct 10, 2026
  5

• Bedrock follow-ups: declined and deferred review findings

  lydakis/agent#103 · Filter by author lydakis opened on Oct 10, 2026
  

• Triggers (T1, #89–#92): review findings we skipped

  lydakis/agent#101 · Filter by author lydakis opened on Oct 10, 2026
  8

• Rich rendering (#81): review findings deferred at merge

  lydakis/agent#97 · Filter by author lydakis opened on Oct 10, 2026
  1