fix(autoname): make Haiku auto-rename reliable and diagnosable#806
Merged
Conversation
Recent Haiku auto-renames failed inconsistently, leaving tasks on their regex slug. Root causes (both confirmed against live ~/.argus logs): - claude -p writes RUNTIME errors (budget exceeded / usage-rate limit / overload) to STDOUT and exits non-zero with an EMPTY stderr. The prior fix folded only ExitError.Stderr, so every real failure logged as a bare "exit status 1" — undiagnosable. - The --max-budget-usd 0.01 cap was tuned to a stale "~$0.0002/call" estimate; a live call measures ~1235 input + ~111 output tokens (~$0.0034), leaving only ~3x headroom. A longer pasted prompt crossed it → "Error: Exceeded USD budget" → exit 1 → slug kept. Fixes in internal/llm/namegen.go: - wrapRunError folds stdout FIRST (the runtime-error channel), then stderr, so failures are diagnosable. - Raise --max-budget-usd 0.01 → 0.05 (~15x measured cost); correct the stale cost comments. - Retry the CLI once with short backoff on a non-zero exit (transient overload/limit); validation failures are NOT retried. - Raise DefaultTimeout 30s → 45s (a signal: killed was seen at 30s). Adds 4 tests (stdout-fold regression, retry-then-succeed, retry-exhausted, budget-flag pin). Routed through openspec (auto-naming) and archived in-PR; gotchas/misc.md updated. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
…lassify Address code-review findings on the retry/diagnostics path: - Log-injection (WARNING): the folded claude stdout reason is untrusted (may echo prompt text) and flowed unescaped into uxlog (%s, newline- terminated) and slog's TextHandler (unquoted). A newline could forge a second physical log line (a fake "[autoname] renamed" record). New scrubReason maps CR/LF/tab/ANSI-ESC/control runes to spaces before folding. - wrapRunError now folds BOTH stdout and stderr when both are present (was: stdout-only, dropping a concurrent flag-parse stderr). - Retry exhaustion kept the LAST attempt's error while the comment claimed "first" — a rich attempt-0 reason (budget) was lost when a retry died bare. Now keeps firstErr; comment corrected; shared-deadline behavior documented honestly. - generateNameOnce returns (name, retryable bool, err) — clearer than the implicit-mutual-exclusion three-error-return. - Dedup double string(out) conversion; tests use testutil.Error. Adds tests: control-char scrub (log-injection guard), keep-first-reason on exhaustion, scrubReason table. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
Merging this branch will not change overall coverage
Coverage by fileChanged files (no unit tests)
Please note that the "Total", "Covered", and "Missed" counts above refer to code statements instead of lines of code. The value in brackets refers to the test coverage of that file in the old version of the code. Changed unit test files
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Recent task auto-renames failed inconsistently, leaving tasks on their
regex slug. Root causes (confirmed against live ~/.argus logs):
overload) to STDOUT and exits non-zero with EMPTY stderr. The prior fix
folded only ExitError.Stderr, so every real failure logged as a bare,
undiagnosable "exit status 1".
estimate; a live call measures ~1235 input +
111 output tokens$0.0034), leaving only ~3x headroom.(
Changes (internal/llm/namegen.go):
stderr when both present, scrubbing control chars so an untrusted
reason cannot forge a second log line (log-injection defense).
keeps the first (richest) reason on exhaustion. DefaultTimeout 30s -> 45s.
Routed through openspec (auto-naming), archived in-PR. Adds 7 tests;
gotchas/misc.md updated.
Co-Authored-By: Claude noreply@anthropic.com