Skip to content

Fix prediction workflow and add X source evidence - #63

Open
kriptoburak wants to merge 1 commit into
XBuilderLAB:mainfrom
kriptoburak:codex/add-tweetclaw-source-evidence
Open

Fix prediction workflow and add X source evidence#63
kriptoburak wants to merge 1 commit into
XBuilderLAB:mainfrom
kriptoburak:codex/add-tweetclaw-source-evidence

Conversation

@kriptoburak

@kriptoburak kriptoburak commented Jul 3, 2026

Copy link
Copy Markdown

Summary

  • add the current TweetClaw install, API key, and OpenClaw allowlist setup
  • map public X results into the existing candidate and benchmark schemas
  • keep post text untrusted and exclude credentials, private data, and drafts
  • preserve Cheat on Content as the scoring and blind-prediction authority

Independent Repository Repairs

Allow valid retrospective edits

The immutability hook previously searched a multiline old_string with grep -F. grep treated every line as a separate pattern. Empty lines matched the protected section, so valid multiline ## 复盘 edits were blocked.

The hook now reconstructs the proposed file and compares protected prediction sections exactly. It:

  • allows metadata and multiline retrospective edits
  • allows a new append-only versioned prediction
  • blocks direct, broad, and replace-all prediction changes
  • fails closed when it cannot verify the edit

A 7-case shell regression suite covers those boundaries.

Finish strict Skill frontmatter compatibility

The recent strict-loader repair missed 4 argument-hint values containing colons. This commit quotes the remaining values in:

  • cheat-learn-from
  • cheat-predict
  • cheat-publish
  • cheat-retro

All root and sub-Skill frontmatter now parses with strict YAML.

Validation

  • bash hooks/prediction-immutability_test.sh (7 passed)
  • LinkedIn extraction checks (7 passed)
  • WeChat adapter tests (3 passed)
  • WeChat runner failure-contract test
  • bash tools/diff_pct_test.sh (3 passed)
  • Bash syntax for every shell script
  • Python compilation for adapters and tools
  • JSON parsing for every tracked JSON file
  • strict YAML parsing for every root and sub-Skill frontmatter block
  • ShellCheck for the changed hook and regression suite
  • GitHub Markdown rendering for every touched document
  • touched-document local-link validation
  • current TweetClaw commands verified against its canonical README
  • git diff --check

The branch is rebuilt as one verified signed commit on current upstream main.

@kriptoburak
kriptoburak force-pushed the codex/add-tweetclaw-source-evidence branch from 25ea5eb to 2e4d623 Compare July 29, 2026 20:47
@kriptoburak kriptoburak changed the title Document optional TweetClaw source evidence Fix prediction workflow and add X source evidence Jul 29, 2026
@kriptoburak

Copy link
Copy Markdown
Author

Rebuilt this PR on current main. Beyond the X source workflow, it now fixes 2 confirmed repository problems: multiline retrospective edits no longer trip the prediction guard, and all 4 remaining colon-bearing argument-hint values load under strict YAML. The new hook suite covers 7 allow/block boundaries, every available repository test passes, and the single commit is verified. Ready for review.

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.

1 participant