Skip to content

Ait

v1.7.1 Breaking

This release includes breaking changes for platform teams planning a safe upgrade.

Published 5d AI Agents & Assistants
✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

agent-harness agent-isolation agentic ai-agents ai-coding ai-tools
+14 more
aider claude-code code-provenance codex coding-assistant cursor developer-tools gemini-cli git git-worktree llm-agents python vcs worktree

Summary

AI summary

Four ait run flags now behave as operators expect and an ENOENT error provides clearer guidance.

Changes in this release

Feature Low

Nested‑wrap warning printed when invoking `ait run` from inside an existing attempt.

Nested‑wrap warning printed when invoking `ait run` from inside an existing attempt.

Source: llm_adapter@2026-05-30

Confidence: high

 —
Bugfix Medium

`--intent <id>` now binds to existing intent when value looks like a ULID.

`--intent <id>` now binds to existing intent when value looks like a ULID.

Source: llm_adapter@2026-05-30

Confidence: high

 —
Bugfix Medium

`--agent <bare-slug>` auto-prepends adapter harness to resolve correctly.

`--agent <bare-slug>` auto-prepends adapter harness to resolve correctly.

Source: llm_adapter@2026-05-30

Confidence: high

 —
Bugfix Medium

`AIT_INTENT` environment variable is now exported alongside `AIT_INTENT_ID`.

`AIT_INTENT` environment variable is now exported alongside `AIT_INTENT_ID`.

Source: llm_adapter@2026-05-30

Confidence: high

 —
Bugfix Medium

ENOENT-127 error message now hints at prompt‑wrap form for prose positional arguments.

ENOENT-127 error message now hints at prompt‑wrap form for prose positional arguments.

Source: llm_adapter@2026-05-30

Confidence: high

 —
Full changelog

Runner-side cascade fixes surfaced by a real-user SLICE_01 reproduction. Four ait run flags that silently did nothing (or produced misleading errors) now work the way operators' mental models expect. Additive only — no breaking changes, no new flags.

Fixed

  • --intent <id> binds to an existing intent when the value looks like a ULID (bare 26-char, intent:<ulid>, or <repo_id>:<ulid>). Previously every value was consumed as the title of a freshly-minted intent — operators who passed an ID got a brand-new intent record with the ULID as its title.
  • --agent <bare-slug> auto-prepends the adapter harness. --adapter claude-code --agent backend-architect now resolves to claude-code:backend-architect instead of failing strict validation.
  • AIT_INTENT env var is now exported alongside AIT_INTENT_ID so the adapter wrapper's shim re-exec path forwards the parent's intent correctly.
  • ENOENT-127 error message hints at the prompt-wrap form when the positional looks like prose. ait run --adapter claude-code "implement..." now suggests ait run [opts] -- claude -p "..." instead of just command not executable.

Added

  • Nested-wrap warning. When ait run --adapter claude-code is invoked from inside an existing attempt (operator tool-calls ait run from a wrapped claude session), AIT prints a stderr warning pointing at AIT_BYPASS=1 and ait off. The shim's own re-exec is exempt via a new AIT_SHIM_REENTRY=1 marker so the warning fires only on operator-typed nested invocations.

See CHANGELOG.md for full detail.

View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track Ait

Get notified when new releases ship.

Sign up free

About Ait

All releases →

Related context

Earlier breaking changes

  • v0.55.52 Keep the built-in claude-code reviewer pinned to the local claude -p CLI even when repository policy defines a conflicting command override.

Beta — feedback welcome: [email protected]