feat(sn60): add improved miner submission kiannidev-20260709-01#107
Closed
kiannidev wants to merge 3 commits into
Closed
feat(sn60): add improved miner submission kiannidev-20260709-01#107kiannidev wants to merge 3 commits into
kiannidev wants to merge 3 commits into
Conversation
carlos4s
added
the
kata:pending
5 tasks
Adds static source probes plus adaptive dual-batch deep audit to beat the current triage-only king within the 3-call inference budget. Co-authored-by: Cursor <cursoragent@cursor.com>
Add import-graph ranking, hot-function triage, risk-compacted audits, Rust discovery, reward-weight static probe, and scorer location hints. Co-authored-by: Cursor <cursoragent@cursor.com>
Rebase on latest kata main and optimize for 2/3 replica pass scoring, smoke-safe execution, and the 150k input-token budget per problem. Co-authored-by: Cursor <cursoragent@cursor.com>
kiannidev
force-pushed
the
feat/kiannidev-20260709-01
branch
from
ae541d5 to
249db1e
Compare
RenzoMXD
added a commit
to RenzoMXD/kata
that referenced
this pull request
Jul 10, 2026
Rewrote the audit strategy after reviewing the public techniques used by other open SN60 submissions (notably PR Autovara#107) for ideas, then implemented everything independently: - Import-graph centrality: file ranking now boosts files many other files import, surfacing hub contracts (routers/vaults) that keyword scoring alone can miss. - LLM triage step: one of the three inference calls now asks the model to pick audit targets from a rich per-file digest (hot functions, state vars, import counts) instead of relying solely on static heuristic ranking for target selection. - Import-linked audit context: deep-audit batches now include a compact excerpt of each target file's direct import dependencies, not just the file in isolation. - Risk-compacted excerpts for oversized files instead of a raw truncation cutoff. - Two zero-cost static probes (unguarded external-call-before-state-write, tx.origin-for-auth) contribute findings before any inference call is spent. - Dedup key now prefers (file, function) over (file, title) so a bug caught by both a probe and the LLM collapses to one reported finding instead of two. Independence check (not just claimed - measured against the actual screening similarity function): - King similarity: 0.094 (review threshold is 0.85) - PR Autovara#107 similarity: 0.083 Validated locally: kata submission validate -> screening_status pass, no review reasons. Synthetic pipeline test (mocked model responses) confirms the triage -> batch -> normalize -> dedupe flow behaves correctly.
Contributor
|
Review re-run by Screening passed. I marked this PR as |
carlos4s
added
kata:executing
Contributor
|
I closed this PR because it failed the round-start execution screener. Your PR passed intake earlier, but before scoring Kata runs the candidate once on a real benchmark project to make sure it can execute and return valid report JSON. Screener project: Reason:
This candidate did not enter the main round scoring, so this is marked |
carlos4s
added
kata:losing
Contributor
|
I closed this PR because your agent did not beat the current king this round. Improve it and open a new PR to try again. How your agent compared to the king this round:
|
3 tasks
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
2 participants
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.
Adds miner bundle
submissions/sn60__bitsec/miner/kiannidev-20260709-01/.Goal
Defend king (
kiannidev-20260708-01) under the July 10, 2026 validation update and beat open competitors.July 2026 rule changes addressed
try/exceptwrapper; fallback file ranking if triage fails{"vulnerabilities": [...]}JSONagent.py(no helper split yet; passes all screening)Pipeline
Self-check
kata submission validate— pass (rebased on latestmain)benchmark_replayscore 0Made with Cursor