Hmbown/CodeWhale · error

Fleet capture cannot resume a session

Error message

Fleet capture cannot resume a session

What it means

When Fleet capture mode is enabled via CODEWHALE_FLEET_CAPTURE_ID and CODEWHALE_FLEET_CAPTURE_DIR, exec_agent runs a fresh captured session. Combining capture with an existing resume_session is contradictory — capture cannot both record a new session and resume an old one — so the run aborts.

Solutions

  1. Unset CODEWHALE_FLEET_CAPTURE_ID and CODEWHALE_FLEET_CAPTURE_DIR when resuming a session.
  2. Drop the --resume flag if the run is intended to be a fresh captured Fleet session.
  3. Scope the capture env vars to the specific fleet invocation rather than exporting them globally.

Example fix

// before
codewhale --resume abc123   # with CODEWHALE_FLEET_CAPTURE_ID set
// after
env -u CODEWHALE_FLEET_CAPTURE_ID -u CODEWHALE_FLEET_CAPTURE_DIR codewhale --resume abc123
Defensive patterns

Strategy: validation

Validate before calling

let capturing = std::env::var_os("CODEWHALE_FLEET_CAPTURE_ID").is_some()
    && std::env::var_os("CODEWHALE_FLEET_CAPTURE_DIR").is_some();
let resuming = resume_session.is_some();
if capturing && resuming {
    return Err(anyhow!("Fleet capture and --resume are mutually exclusive"));
}

Try / catch

match run_exec_agent(args).await {
    Err(e) if e.to_string().contains("Fleet capture cannot resume") => {
        eprintln!("disable Fleet capture env vars or drop --resume");
        std::process::exit(2);
    }
    other => other?,
}

Prevention

When it happens

Trigger: Setting both CODEWHALE_FLEET_CAPTURE_ID/CODEWHALE_FLEET_CAPTURE_DIR in the environment AND passing a resume session (e.g. `--resume <id>` or an equivalent programmatic resume_session value) into run_exec_agent.

Common situations: A fleet orchestration script exports capture env vars globally while a developer also runs a manual resume command in the same shell; CI inherits capture variables from a parent environment and then performs session resume tests.

Related errors

AI-assisted analysis of Hmbown/CodeWhale@73e0f67d83 (2026-09-22). Data as JSON: /api/errors/a5ece062a00e6165. Report an issue: GitHub.

Appendix: source

Thrown at crates/tui/src/exec_agent.rs:347

    let fleet_authority_active = fleet_authority.is_some();
    let outer_network_access = fleet_authority
        .as_ref()
        .and_then(|authority| authority.network_access);
    let outer_shell_authority = fleet_authority
        .as_ref()
        .map(|authority| authority.shell)
        .unwrap_or_default();
    if let Some(envelope) = fleet_authority {
        crate::tools::spec::install_process_tool_authority(envelope).map_err(anyhow::Error::msg)?;
    }

    let fleet_capture = match (
        std::env::var("CODEWHALE_FLEET_CAPTURE_ID").ok(),
        std::env::var_os("CODEWHALE_FLEET_CAPTURE_DIR"),
    ) {
        (Some(id), Some(dir)) => {
            uuid::Uuid::parse_str(&id).context("invalid Fleet session capture id")?;
            anyhow::ensure!(
                resume_session.is_none(),
                "Fleet capture cannot resume a session"
            );
            let manager = SessionManager::new(PathBuf::from(dir))?;
            match manager.load_session(&id) {
                Err(err) if err.kind() == std::io::ErrorKind::NotFound => {}
                _ => anyhow::bail!("Fleet session capture id is already in use or unavailable"),
            }
            Some((id, manager))
        }
        (None, None) => None,
        _ => anyhow::bail!("incomplete Fleet session capture destination"),
    };

    let route = resolve_cli_exec_route(config, model, prompt, force_configured_route).await?;
    let execution_config = config_for_cli_route(config, &route);
    let auto_model = route.auto_model;
    let effective_provider = route.provider;

View on GitHub (pinned to 73e0f67d83)