A refusal by a guard counts as a failure, and an edit is generated by V3 before its syntax is checked
Maintainers usually reply within 1 day
Nobody has claimed this yet.
Assessment
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Newbie friendliness
- 45/100
- Issue type
- Bug
- Clarity
- Mostly clear
- Activity status
- Active
- Tech stack
- python
- Domain
- ai-infra-agents
Research direction
Start by resolving the dependency on #326, then trace the guard refusal counters, the four edit tools, and the syntax gate; the issue does not name files or tests. Add tests showing that one failure plus two different refusals continue the session and that syntax errors are rejected before V3 runs. Done means the ending reason is also shown in the TUI and counted by the scorers.
Written by the indexing model from the issue text.
Description
What happens
- When one of ATLAS's own guards refuses a tool call (an existing
file, an argument name, a permission, the command policy), the
refusal is added to the same counters as a real failure. One real
failure plus two different refusals stop the session, and the
ending message says the same call was re-sent unchanged, which is
not true. - The four edit tools call V3 first and run the syntax gate after
it. When the gate refuses, the V3 work is thrown away.write_file
already checks the syntax first.
What needs doing
- A separate counter for refusals, with its own limits and its own
ending reason. An identical re-send still stops early. A reply that
cannot be parsed gets its own streak. - The syntax check runs before V3 for edits, as it does for whole
files.
Done when
- A session with one failure and two different refusals goes on.
- An edit with a syntax error is refused before V3 runs.
- Tests for both fail on the old behaviour.
- The new ending reason is shown in the TUI and counted by the
scorers.
Blocked by: #326
- Dominant language
- Python
- Stars
- 2.1k
- Forks
- 197
- Avg merge
- 14h 1m
- Merged PRs (30d)
- 201
Getting set up
- Ships a Dockerfile or Docker Compose file
- Has a pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from inferstep/ATLAS
-
Bootstrap: the last two steps run as the user who owns the installMay be free again @assassinationss claimed this 19 days ago, and no pull request is open. Openarea/install reliability
Difficulty 2/5 1-3 hours Newbie friendliness 75/100
inferstep/ATLAS#412 · 4 comments ·
Maintainers usually reply within 1 day
-
area/cli
Difficulty 2/5 1-3 hours Newbie friendliness 85/100
Maintainers usually reply within 1 day
-
area/cli area/lens status/ready
Difficulty 2/5 Half a day Newbie friendliness 66/100
Maintainers usually reply within 1 day
-
area/ci status/ready
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
inferstep/ATLAS#245 · 4 comments ·
Maintainers usually reply within 1 day
-
area/docs documentation help wanted
Difficulty 4/5 3-5 days Newbie friendliness 55/100
Maintainers usually reply within 1 day
Similar issues
-
enhancement
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
glossyio/traffic-monitor#264 ·
Maintainers usually reply within 1 day
-
Add self-hostOpen
Difficulty 1/5 Under an hour Newbie friendliness 75/100
Atarity/deploy-your-own-saas#102 ·
-
[Bug]: RDFlex silently drops the covariate adjustment when the outcome or treatment column is integer-typedPossibly taken @vrjdev claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
DoubleML/doubleml-for-py#424 ·
Maintainers usually reply within 1 day
-
agentic-dmx docs enhancement
Difficulty 1/5 1-3 hours Newbie friendliness 86/100
-
Difficulty 1/5 Under an hour Newbie friendliness 72/100
521xueweihan/HelloGitHub#4150 ·