Skip to content

docs(rf-core): add failure fingerprints to debug - #14

Merged
manykarim merged 4 commits into
MarketSquare:mainfrom
ai94iq:docs/failure-fingerprints
Nov 17, 2026
Merged

manykarim merged 4 commits into
MarketSquare:mainfrom
ai94iq:docs/failure-fingerprints

Conversation

@ai94iq

@ai94iq ai94iq commented Nov 15, 2026 •

Copy link
Copy Markdown
Contributor

Adds failure fingerprints to the rf-debug-expert subagent.

Verified: build-agent-variants --check passes (7 variants up to date). Catalog, governance, plugin references and docs voice checks pass.

What this changes

Groups failures by a fingerprint of the error message, so the same failure in a later run matches a known cause. Says RequestsLibrary puts the actual value first in its messages.

Review checklist

Tick each item, or write "n/a: " after it. Reviewers don't approve a PR
with an unticked item that has no reason (see CONTRIBUTING.md, "Review").

  • Purpose. Each new or changed skill and subagent says what it is for and when to use it, in its description. n/a: the subagent description is unchanged
  • Supported versions. catalog.yaml lists the libraries and version ranges the content was checked against, and the Robot Framework range. n/a: no new library, catalog.yaml is unchanged
  • Examples. .robot examples pass robot --dryrun; Python examples run. n/a: no examples
  • Security-relevant changes. Hooks, scripts, MCP configuration, manifests, subagent tools and skill allowed-tools changes are listed below; the security gate passes, and any new baseline entry has a reason. n/a: subagent text only, its tools list is unchanged
  • Tool permissions. Every subagent declares an explicit tools list; no allowed-tools entry grants unrestricted shell access. n/a: tools list unchanged
  • Documentation voice. Text follows the writing rules in CONTRIBUTING.md and passes the docs voice check.

Security-relevant changes

none

Add a fingerprint step and a rule-out checklist to rf-debug-expert.
The agent names each failure group by a stable signature and checks
the test before it reports an app or environment cause. The bundle
and the agent variants are regenerated from the plugin.
@ai94iq
ai94iq requested a review from manykarim as a code owner November 15, 2026 07:47
Section 6 now says RequestsLibrary puts the actual value first and that
other libraries order it differently. Section 2 now says the fingerprint
goes in the bug title or BUGS.md entry. Same text in all four copies.

Checked: cmp of plugin and bundle copies is identical; grep finds no bare
python scripts/ command. scripts/check-drift.sh could not run: Git Bash
python is the Windows Store alias (log ../drift.log).
@ai94iq

ai94iq commented Nov 15, 2026

Copy link
Copy Markdown
Contributor Author

Merge chain: #11, then #12, then #13, then #14, then #15, then #16.

All six merge into main cleanly in this order (dry run on 2026-11-15).

Apart from #13 and #15, the PRs do not depend on each other.

manykarim and others added 2 commits November 17, 2026 14:14
BUGS.md isn't a common convention, so the agent might start creating
one. Regenerated with python scripts/build-agent-variants.py and
python scripts/build-catalog.py.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FHEmzL3ErAhGQykKWDx2wP
@manykarim
manykarim merged commit 21c8301 into MarketSquare:main Nov 17, 2026
26 checks passed
@manykarim

Copy link
Copy Markdown
Collaborator

Thanks a lot, @ai94iq! 🎉 The "rule out the test first" checklist is a great addition to rf-debug-expert, and the RequestsLibrary point is right: in 400 != 200 the first value is what the server sent.

I made one small change so I could merge it right away:

  • merged the current main into your branch;
  • changed "the bug title or the BUGS.md entry" to "the bug title or report" (plugins/rf-core/agents/rf-debug-expert.md). BUGS.md isn't a common convention in Robot Framework projects, so an agent might otherwise start creating one;
  • regenerated the agent variants and the bundle with python scripts/build-agent-variants.py and python scripts/build-catalog.py.

All checks passed, and it's merged now.

One small note for future PRs: since you opened this, we published AI_POLICY.md, and the PR template now has an "AI assistance" section. On new PRs we'd love a line there whenever a tool did a substantial part of the work, plus the command that regenerated bundles/.

Thanks again!

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.

2 participants