Skip to content

fix(tools): clarify kanban_complete phantom-card retry guidance (salvage #23072) - #23453

Merged
teknium1 merged 1 commit into
mainfrom
salvage/pr-23072-phantom-retry-guidance
May 10, 2026
Merged

fix(tools): clarify kanban_complete phantom-card retry guidance (salvage #23072)#23453
teknium1 merged 1 commit into
mainfrom
salvage/pr-23072-phantom-retry-guidance

Conversation

@teknium1

Copy link
Copy Markdown
Contributor

Summary

Workers no longer abandon a kanban task when kanban_complete rejects a hallucinated created_cards claim. The HallucinatedCardsError tool_error message now explicitly states that the task is still in-flight and tells the worker how to retry — either with a corrected created_cards list or with created_cards=[] to skip the card-claim check entirely.

Root cause

The kanban kernel's hallucinated-card gate already left the task runnable on phantom rejection (no state change, the worker can call kanban_complete again). But the previous tool_error message read like a terminal failure, so workers were observed abandoning the run, tripping the consecutive_failures breaker, and ending up auto-blocked. Pure messaging issue — kernel behavior was correct, the recovery path just wasn't legible to the model.

Changes

  • tools/kanban_tools.py: rewrote HallucinatedCardsError's tool_error to lead with "Your task is still in-flight (no state change). Retry kanban_complete with ... created_cards=[] to skip the card-claim check entirely." Workers can now self-correct without the operator having to intervene.
  • tests/tools/test_kanban_tools.py: 3 new tests — pin the literal phrasing the worker depends on to discover the retry path, plus successful retry via created_cards=[] and via a corrected list.
  • tests/hermes_cli/test_kanban_core_functionality.py: 1 new kernel-level test (test_complete_can_retry_after_phantom_rejection) exercising both retry shapes and asserting the audit + final completed event land.

No kernel behavior change — just messaging plus regression tests that lock the recovery cues so a future reword can't silently break worker self-recovery.

Validation

Before After
tests/tools/test_kanban_tools.py + test_kanban_core_functionality.py (phantom subset) 0/0 4/4
tests/tools/test_kanban_tools.py + test_kanban_core_functionality.py + test_kanban_db.py 290/290 294/294

Closes #22923 via salvage. Salvage of #23072. Original commit by @konsisumer cherry-picked with authorship preserved (re-attributed during salvage from der@konsi.org to the GitHub-noreply form for release-notes credit). AUTHOR_MAP entry already exists from yesterday's #23071 salvage.

When kanban_complete rejects a created_cards list as hallucinated, the
task is intentionally left in-flight (the gate runs before the write
txn) so the worker can retry with a corrected list or pass
created_cards=[] to skip the check. The retry path already worked, but
the previous error wording read like a terminal failure and workers
were observed abandoning the run instead of trying again.

Spell out the recovery path explicitly in the tool_error response
("Your task is still in-flight ... Retry kanban_complete with ...") and
add regression coverage at both the kernel and tool layers so the
retry contract — and the wording the worker depends on to discover
it — is pinned.

Fixes #22923
@teknium1
teknium1 merged commit 62cfe79 into main May 10, 2026
12 of 15 checks passed
@teknium1
teknium1 deleted the salvage/pr-23072-phantom-retry-guidance branch May 10, 2026 23:14
@github-actions

Copy link
Copy Markdown
Contributor

🔎 Lint report: salvage/pr-23072-phantom-retry-guidance vs origin/main

ruff

Total: 0 on HEAD, 0 on base (➖ 0)

🆕 New issues: none

✅ Fixed issues: none

Unchanged: 0 pre-existing issues carried over.

ty (type checker)

Total: 8085 on HEAD, 8060 on base (🆕 +25)

🆕 New issues (15):

Rule Count
invalid-argument-type 8
unsupported-operator 3
unresolved-attribute 3
not-subscriptable 1
First entries
tests/agent/test_codex_cloudflare_headers.py:181: [unsupported-operator] unsupported-operator: Operator `in` is not supported between objects of type `Literal["originator"]` and `(Unknown & ~AlwaysFalsy) | (str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | ... omitted 3 union elements`
run_agent.py:2641: [invalid-argument-type] invalid-argument-type: Argument to function `get_model_context_length` is incorrect: Expected `str`, found `str | dict[str, str] | Any | ... omitted 3 union elements`
tests/agent/test_codex_cloudflare_headers.py:163: [unresolved-attribute] unresolved-attribute: Attribute `startswith` is not defined on `dict[str, str]` in union `Unknown | str | dict[str, str]`
tests/run_agent/test_provider_attribution_headers.py:155: [unsupported-operator] unsupported-operator: Operator `not in` is not supported between objects of type `Literal["X-OpenRouter-Cache"]` and `Unknown | str | dict[str, str] | ... omitted 3 union elements`
run_agent.py:13287: [invalid-argument-type] invalid-argument-type: Argument to function `len` is incorrect: Expected `Sized`, found `(str & ~AlwaysFalsy) | (dict[Unknown, Unknown] & ~AlwaysFalsy) | (Any & ~AlwaysFalsy) | ... omitted 3 union elements`
tests/run_agent/test_provider_attribution_headers.py:90: [unresolved-attribute] unresolved-attribute: Attribute `startswith` is not defined on `dict[str, str]` in union `Unknown | str | dict[str, str]`
run_agent.py:2339: [invalid-argument-type] invalid-argument-type: Argument to function `query_ollama_num_ctx` is incorrect: Expected `str`, found `(str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | (Any & ~AlwaysFalsy) | ... omitted 4 union elements`
run_agent.py:7160: [invalid-argument-type] invalid-argument-type: Argument to function `build_anthropic_client` is incorrect: Expected `str`, found `str | dict[Unknown, Unknown] | Any | ... omitted 3 union elements`
run_agent.py:2593: [invalid-argument-type] invalid-argument-type: Argument to function `_is_oauth_token` is incorrect: Expected `str`, found `(Unknown & ~AlwaysFalsy) | (str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | ... omitted 4 union elements`
run_agent.py:2590: [invalid-argument-type] invalid-argument-type: Argument to function `build_anthropic_client` is incorrect: Expected `str`, found `(Unknown & ~AlwaysFalsy) | (str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | ... omitted 4 union elements`
run_agent.py:13284: [invalid-argument-type] invalid-argument-type: Argument to function `_is_oauth_token` is incorrect: Expected `str`, found `str | dict[Unknown, Unknown] | Any | ... omitted 3 union elements`
run_agent.py:6989: [invalid-argument-type] invalid-argument-type: Argument to function `_codex_cloudflare_headers` is incorrect: Expected `str`, found `Unknown | str | dict[str, str] | ... omitted 3 union elements`
tests/agent/test_codex_cloudflare_headers.py:163: [unresolved-attribute] unresolved-attribute: Attribute `get` is not defined on `str & ~AlwaysFalsy`, `int & ~AlwaysFalsy` in union `(Unknown & ~AlwaysFalsy) | (str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | ... omitted 3 union elements`
tests/run_agent/test_provider_attribution_headers.py:156: [unsupported-operator] unsupported-operator: Operator `not in` is not supported between objects of type `Literal["X-OpenRouter-Cache-TTL"]` and `Unknown | str | dict[str, str] | ... omitted 3 union elements`
tests/run_agent/test_provider_attribution_headers.py:154: [not-subscriptable] not-subscriptable: Cannot subscript object of type `int` with no `__getitem__` method

✅ Fixed issues (9):

Rule Count
invalid-argument-type 8
unresolved-attribute 1
First entries
run_agent.py:7160: [invalid-argument-type] invalid-argument-type: Argument to function `build_anthropic_client` is incorrect: Expected `str`, found `str | dict[Unknown, Unknown] | Any | ... omitted 4 union elements`
tests/agent/test_codex_cloudflare_headers.py:163: [unresolved-attribute] unresolved-attribute: Attribute `get` is not defined on `str & ~AlwaysFalsy` in union `(Unknown & ~AlwaysFalsy) | (str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | dict[Unknown, Unknown] | Divergent`
run_agent.py:2641: [invalid-argument-type] invalid-argument-type: Argument to function `get_model_context_length` is incorrect: Expected `str`, found `str | dict[str, str] | Any | ... omitted 4 union elements`
run_agent.py:2590: [invalid-argument-type] invalid-argument-type: Argument to function `build_anthropic_client` is incorrect: Expected `str`, found `(Unknown & ~AlwaysFalsy) | (str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | ... omitted 5 union elements`
run_agent.py:13284: [invalid-argument-type] invalid-argument-type: Argument to function `_is_oauth_token` is incorrect: Expected `str`, found `str | dict[Unknown, Unknown] | Any | ... omitted 4 union elements`
run_agent.py:2339: [invalid-argument-type] invalid-argument-type: Argument to function `query_ollama_num_ctx` is incorrect: Expected `str`, found `(str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | (Any & ~AlwaysFalsy) | ... omitted 5 union elements`
run_agent.py:6989: [invalid-argument-type] invalid-argument-type: Argument to function `_codex_cloudflare_headers` is incorrect: Expected `str`, found `Unknown | str | dict[str, str] | dict[Unknown, Unknown] | Divergent`
run_agent.py:13287: [invalid-argument-type] invalid-argument-type: Argument to function `len` is incorrect: Expected `Sized`, found `(str & ~AlwaysFalsy) | (dict[Unknown, Unknown] & ~AlwaysFalsy) | (Any & ~AlwaysFalsy) | ... omitted 4 union elements`
run_agent.py:2593: [invalid-argument-type] invalid-argument-type: Argument to function `_is_oauth_token` is incorrect: Expected `str`, found `(Unknown & ~AlwaysFalsy) | (str & ~AlwaysFalsy) | (dict[str, str] & ~AlwaysFalsy) | ... omitted 5 union elements`

Unchanged: 4240 pre-existing issues carried over.

Diagnostics are surfaced as warnings — this check never fails the build.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[Bug]: kanban_complete hallucinated-card gate blocks completion with no recovery — worker must crash or abandon task

2 participants