feat(cli): add support for NO_DNA agent mode (full no-dna.org complia…

[rit3sh-x]

Summary

Detect NO_DNA per https://no-dna.org and adapt CLI output for AI agents and
automation. Full compliance with all 6 no-dna.org behavioral bullets.

Under NO_DNA surfpool:

• skips interactive prompts (no_tui, skip_runbook_generation_prompts,
  unsupervised, skip_confirm forced)
• never constructs spinners or progress bars
• routes log output to stderr as JSONL with RFC3339 UTC timestamps
• disables ANSI color (NO_COLOR honored independently)
• emits one JSON-per-runbook on surfpool ls
• composes cleanly with --ci (orthogonal)

Test plan

• cargo test -p surfpool-cli --bins — 50/50 passing
• cargo +nightly fmt --all clean
• cargo check -p surfpool-cli clean
• Manual: NO_DNA=1 surfpool start --offline → stdout clean, JSON lines on stderr
• Manual: NO_DNA=1 surfpool ls → JSON-per-runbook on stderr
• Manual: surfpool start (no NO_DNA) → unchanged human behavior
• Manual: NO_DNA=1 surfpool start --ci → both flag sets apply

Closes #569. Supersedes #593.

[greptile-apps]

Greptile Summary

This PR adds full NO_DNA agent-mode support to the surfpool CLI, implementing the no-dna.org spec so that setting NO_DNA to any non-empty value switches the binary to a machine-friendly mode.

• no_dna.rs + agent_io.rs: New modules detect NO_DNA/NO_COLOR once per process via OnceLock, expose an AgentMode value struct, and provide cli_emit/agent_emit helpers plus cli_info!/cli_warn!/cli_error! macros that route to either human stdout/stderr or JSONL on stderr.
• Behavioral changes: Color macros gain an AgentMode guard to prevent ANSI leakage; persist_log suppresses its direct-print path under agent mode to avoid duplicate stderr lines; handle_log_event short-circuits transient events into persist_log (no spinners); the update progress bar is skipped in favour of cli_info! markers; surfpool ls emits one JSON object per runbook.
• --ci orthogonality: The old inline if cmd.runtime.ci block is extracted into apply_ci_mode_to_simnet, sitting alongside apply_agent_mode_to_simnet so both flag-sets compose independently, with dedicated tests confirming neither silently suppresses the other.

Confidence Score: 5/5

This PR is safe to merge; no changed code path introduces incorrect behavior in human or agent mode.

The NO_DNA detection and routing logic is cleanly separated and well-tested. The OnceLock ensures consistent state across all call sites. The color macro guard correctly prevents ANSI leakage before strip_ansi is even needed. The fern dispatch routing and suppress_cli logic in persist_log are correct and avoid double-emitting. No pre-existing behavior is regressed for human-mode users.

agent_io.rs — the strip_ansi helper leaves non-CSI escape sequence bodies in the output, which is a minor gap given the libraries in use only emit CSI sequences.

Important Files Changed

Filename	Overview
crates/cli/src/no_dna.rs	New file implementing AgentMode detection from NO_DNA/NO_COLOR env vars using OnceLock caching; well-tested with from_raw/from_env split.
crates/cli/src/agent_io.rs	New file implementing JSONL emit for agent mode; strip_ansi helper handles CSI sequences correctly but leaves OSC/non-CSI escape sequence bodies intact.
crates/cli/src/macros.rs	Color macros now gate on AgentMode::from_env().color != Never in addition to the TTY check, preventing ANSI leak into JSON msg fields.
crates/cli/src/cli/mod.rs	setup_logger gains agent_mode param; agent-mode helpers extracted for simnet/run/update; handle_list_command emits per-runbook JSON under NO_DNA; extensive tests added.
crates/cli/src/runbook/mod.rs	persist_log suppresses cli_* emit under agent mode to avoid double stderr; handle_log_event routes transient events to persist_log (no spinners) in agent mode.
crates/cli/src/cli/update/mod.rs	Progress bar skipped under NO_DNA; replaced with cli_info! start/end markers; println!/eprintln! sites migrated to cli_info!/cli_warn! macros.
crates/cli/src/cli/simnet/mod.rs	Single eprintln! replaced with cli_error!; comment added explaining spinners are never populated under agent mode.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    Start([surfpool invoked]) --> Init[AgentMode from_env - OnceLock cached]
    Init --> Check{NO_DNA set and non-empty?}
    Check -- No --> Human[Human mode - Human format, Auto color]
    Check -- Yes --> Agent[Agent mode - JSON format, Never color]
    Human --> HumanLog[fern dispatch: console to stdout, colored + local timestamp]
    Human --> HumanSpinner[ProgressBar and spinners enabled]
    Agent --> AgentLog[fern dispatch: console to stderr, JSONL + RFC3339 UTC]
    Agent --> AgentPrint[cli_emit macros: agent_emit to stderr]
    Agent --> AgentNoSpinner[ProgressBar skipped, no spinners]
    AgentPrint --> StripANSI[strip_ansi defense-in-depth]
    StripANSI --> JsonOut[JSONL record on stderr]
    AgentLog --> JsonOut
    HumanLog --> HumanOut[colored lines on stdout or stderr]

Loading

_{Reviews (5): Last reviewed commit: "fix(cli): dedupe NO_DNA JSONL emission f..." | Re-trigger Greptile}

[rit3sh-x]

@lgalabru @MicaiahReid
Closes #569, supersedes #593. Full no-dna.org compliance (all 6 bullets, 50/50 tests).

Deliberately left for follow-ups: static enforcement of NO_DNA gating (every invariant is by-convention today one careless PR away from regression), version field in the JSON contract (add before agents pin to it), and stabilizing the target field (currently a Rust module path, refactor breaks parsers).

Lower priority: structured data payloads on more sites, MCP JSONL, NO_COLOR in TUI palettes, CHANGELOG/README entry. Happy to file these as issues after merge.

[rit3sh-x]

@lgalabru @MicaiahReid does the implementation look good, or would you like to steer the direction?

[MicaiahReid]

@lgalabru @MicaiahReid does the implementation look good, or would you like to steer the direction?

@rit3sh-x we'll need some patience on this review 🙏
We'll let you know when we have comments