fix: read tool_response instead of tool_output in PostToolUse hook

[faraz152]

Root cause

src/hooks/post-tool-use.ts (and its compiled counterpart plugin/scripts/post-tool-use.mjs) read the tool output from data.tool_output:

const { imageData, cleanOutput } = extractImageData(data.tool_output);

But Claude Code's PostToolUse hook payload uses tool_response, not tool_output. This means cleanOutput is always undefined, the observe request reaches the server with no output value, and mem::compress fails schema validation on every real tool call — narrative and facts are empty so the XML schema check fails.

Captured real PostToolUse payload from Claude Code (from #539):

{
  "hook_event_name": "PostToolUse",
  "tool_name": "Bash",
  "tool_input": { "command": "echo hi" },
  "tool_response": { "stdout": "hi\n", "stderr": "", "interrupted": false }
}

Fix

Read data.tool_response with data.tool_output as a fallback so any older integration still using the legacy field name keeps working:

const { imageData, cleanOutput } = extractImageData(
  data.tool_response ?? data.tool_output,
);

Same one-liner applied to the compiled plugin/scripts/post-tool-use.mjs.

Impact

Without this fix mem::compress silently fails on ~100% of real Claude Code tool calls, degrading memory_recall output and starving the consolidation pipeline of compressed observations.

Fixes #539

Summary by CodeRabbit

• Bug Fixes
  • Improved tool response processing to prefer the newer response field when present, falling back to legacy output otherwise.
  • Ensures image data and cleaned output are correctly extracted and included in downstream observations, improving compatibility and reliability across response formats.

[vercel]

@faraz152 is attempting to deploy a commit to the rohitg00's projects Team on Vercel.

A member of the Team first needs to authorize it.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: b932002d-5979-44c7-96d3-59c93a65b8b6

📥 Commits

Reviewing files that changed from the base of the PR and between bb51ec0 and d0dc2ac.

📒 Files selected for processing (1)

• src/hooks/post-tool-use.ts

💤 Files with no reviewable changes (1)

• src/hooks/post-tool-use.ts

---

📝 Walkthrough

Walkthrough

Both the TypeScript hook and JavaScript script now use data.tool_response ?? data.tool_output when extracting image data and cleaning tool output, ensuring the code reads Claude Code's tool_response field while preserving legacy compatibility.

Changes

Tool Response Compatibility

Layer / File(s)	Summary
Tool response field fallback with backward compatibility src/hooks/post-tool-use.ts, plugin/scripts/post-tool-use.mjs	Both files now use data.tool_response ?? data.tool_output as the input to extractImageData, switching which field is scanned and sanitized for embedded/base64 image data while keeping backward compatibility with legacy payloads.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Poem

A rabbit finds a field named two,
response first, then output too.
It sniffs for pictures, cleans the rest,
Hops on quietly — backward-compatible best.
🐇✨

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and specifically describes the main change: fixing the PostToolUse hook to read tool_response instead of tool_output, which directly addresses the primary bug in the PR.
Linked Issues check	✅ Passed	The PR fully satisfies issue #539 by implementing the proposed patch to use data.tool_response with data.tool_output as fallback in both source and compiled files, fixing the mem::compress failure.
Out of Scope Changes check	✅ Passed	All changes are directly scoped to fixing the tool_response/tool_output field-name mismatch in the PostToolUse hook; no unrelated modifications are present.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

🧹 Nitpick comments (1)

src/hooks/post-tool-use.ts (1)

35-36: ⚡ Quick win

Remove the explanatory WHAT comment and rely on clear code instead.

The fallback expression is already self-explanatory; please drop the inline explanatory comment to align with repository style rules.

As per coding guidelines, "Avoid code comments explaining WHAT — use clear naming instead".

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/hooks/post-tool-use.ts` around lines 35 - 36, Remove the explanatory
inline comment and rely on the clear fallback expression itself: delete the
comment line that describes "Claude Code's PostToolUse payload uses
`tool_response`; fall back to `tool_output`…" and keep the existing fallback
expression that references `tool_response` and `tool_output` unchanged so the
code reads cleanly without a WHAT-style comment.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@src/hooks/post-tool-use.ts`:
- Around line 35-36: Remove the explanatory inline comment and rely on the clear
fallback expression itself: delete the comment line that describes "Claude
Code's PostToolUse payload uses `tool_response`; fall back to `tool_output`…"
and keep the existing fallback expression that references `tool_response` and
`tool_output` unchanged so the code reads cleanly without a WHAT-style comment.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 035c737b-679f-494d-a29e-af86185049f8

📥 Commits

Reviewing files that changed from the base of the PR and between 1838f4d and bb51ec0.

📒 Files selected for processing (2)

• plugin/scripts/post-tool-use.mjs
• src/hooks/post-tool-use.ts

[rohitg00]

Thanks @faraz152 — clean root-cause writeup with a captured payload, exactly the kind of repro that makes review easy. Merged.