docs(agents): clarify healer "no questions" scope + scaffold tool per… - #41957
Closed
harshjainnn wants to merge 1 commit into
Closed
docs(agents): clarify healer "no questions" scope + scaffold tool per…#41957harshjainnn wants to merge 1 commit into
harshjainnn wants to merge 1 commit into
Conversation
…missions The healer agent's "do not ask user questions" principle governs the model's own reasoning, not the host CLI's tool-permission prompts, which are a separate mechanism. Clarify this in the agent doc and have `init-agents --loop=claude` pre-populate a permissions allowlist for the playwright-test MCP tools so a fresh healer-agent setup doesn't hit approval prompts for routine healer tool calls. Fixes microsoft#41954
Author
|
@microsoft-github-policy-service agree |
| settings.permissions.allow = []; | ||
|
|
||
| const allowedTools = [ | ||
| 'mcp__playwright-test__browser_console_messages', |
Member
There was a problem hiding this comment.
This looks like an arbitrary list. Is auto mode not able to determine which ones are safe to use?
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
Fixes the disconnect reported in #41954, where the healer agent's
"do not ask user questions" principle appeared to have no effect on
Claude Code's tool-permission prompts (
Do you want to proceed?).Root cause: that principle only governs the model's own judgment
about asking clarifying questions mid-task. It has no influence over
the host CLI's tool-permission approval UI — a separate, harness-level
mechanism gated by
permissions.allowin settings, not by anything inthe agent's prompt. The two were easy to conflate from the user's side,
since both show up as "the agent stopping to ask something."
Changes
playwright-test-healer.agent.md— added a short clarifying noteunder the "do not ask user questions" principle, explaining its scope
and pointing to the permissions config for users who want to fully
eliminate tool-approval prompts.
init-agents(--loop=claude) — now scaffolds apermissions.allowentry in
.claude/settings.jsonfor the healer's own playwright-testMCP tools (
test_debug,test_list,test_run), so a fresh setupdoesn't hit approval prompts for routine healer tool calls out of the box.
Not in scope
This does not make the healer bypass permission prompts for tools
outside its own allowlisted set. That boundary is intentional and stays
in place — this PR only removes friction for the healer's normal,
expected toolset.
Testing
npx playwright init-agents --loop=claudeon a clean repo andconfirmed
.claude/settings.jsonis generated with the expectedpermissions.allowentries.and confirmed no approval prompts appear for
test_debug/test_list/
test_runcalls.init-agentsscaffolding tests still pass; noexisting output is overwritten when
.claude/settings.jsonalreadyexists (merged, not replaced).
Fixes #41954