feat: improve Slack notification formatting + read issue comments#1238
feat: improve Slack notification formatting + read issue comments#1238Gkrumbach07 merged 6 commits intomainfrom
Conversation
The issue body may be sparse with important context in comments. Explicitly instruct the agent to use `gh issue view --comments`. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Use Slack mrkdwn link format (<URL|text>) instead of raw URLs - Include issue/PR title in notifications - PR created notification includes a brief summary of changes - Question notification shows the actual question prominently - Need-help notification explains what was tried and why stuck Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
📝 WalkthroughWalkthroughWorkflow now sends Slack notifications before AskUserQuestion, dynamically builds mrkdwn Slack payloads by fetching issue/PR titles via Changes
Sequence Diagram(s)sequenceDiagram
participant Workflow
participant GH_CLI as "GitHub CLI"
participant Payload as "jq (payload builder)"
participant Slack as "Slack Webhook"
participant Agent as "Agent / AskUserQuestion"
Workflow->>GH_CLI: gh pr/issue view --json title
GH_CLI-->>Workflow: title JSON
Workflow->>Payload: build Slack JSON (jq -nc, mrkdwn links, emojis, session link)
Payload-->>Workflow: assembled payload
Workflow->>Slack: POST notification (send BEFORE AskUserQuestion)
Slack-->>Workflow: 200 OK
Workflow->>Agent: AskUserQuestion
Note right of Workflow: on PR creation
Workflow->>Payload: build "PR created" payload (PR link + session link + summary)
Payload-->>Workflow: PR payload
Workflow->>Slack: POST "PR created" notification
Slack-->>Workflow: 200 OK
Important Pre-merge checks failedPlease resolve all errors before merging. Addressing warnings is optional. ❌ Failed checks (2 errors)
✅ Passed checks (4 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
✨ Simplify code
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (1)
.github/workflows/amber-issue-handler.yml (1)
85-86:gh issue view --commentsis required, but session auth/context is not guaranteed.These new instructions make
ghmandatory, but the ambient session env in this workflow doesn’t currently passGH_TOKEN. Ifghisn’t pre-authenticated in the agent runtime, this step will fail and block the intended behavior.Suggested hardening
environment-variables: >- - {"SLACK_WEBHOOK_URL": "${{ secrets.SLACK_WEBHOOK_URL }}", "PLATFORM_HOST": "${{ secrets.PLATFORM_HOST }}"} + {"SLACK_WEBHOOK_URL": "${{ secrets.SLACK_WEBHOOK_URL }}", "PLATFORM_HOST": "${{ secrets.PLATFORM_HOST }}", "GH_TOKEN": "${{ secrets.GITHUB_TOKEN }}"}Apply similarly to both issue-fix ambient-action steps.
Also applies to: 281-282
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.github/workflows/amber-issue-handler.yml around lines 85 - 86, The workflow requires running the CLI command `gh issue view --comments` but does not guarantee `gh` is authenticated; update the two "issue-fix ambient-action" steps that call `gh issue view --comments` to ensure authentication by either 1) adding a preceding auth/setup step (e.g., use an `actions/setup-gh` or run `gh auth login --with-token` with a provided secret), or 2) conditionally checking `gh auth status` and falling back to using the REST API with `GITHUB_TOKEN`/`secrets.GH_TOKEN`; ensure the `GH_TOKEN` (or `GITHUB_TOKEN`) secret is passed into the job environment and referenced when authenticating so the `gh` invocation cannot fail due to missing session auth.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In @.github/workflows/amber-issue-handler.yml:
- Around line 291-292: The Slack notification payloads only include the issue/PR
number (steps.context.outputs.number) and omit the title; update the message
templates that build the --data JSON (the lines using steps.context.outputs.url
and steps.context.outputs.number) to also interpolate the issue/PR title from
steps.context.outputs.title (or the equivalent output produced by the context
step) into the message text (e.g., "❓ *Question about <${{
steps.context.outputs.url }}|#${{ steps.context.outputs.number }} - ${{
steps.context.outputs.title }}>*"), and apply the same change to the other
similar templates referenced (the PR/comment question and PR notification
templates) so every Slack payload includes both number and title. Ensure proper
escaping/quoting in the JSON string when inserting the title.
- Around line 95-96: The Slack payload embeds ${{ steps.issue.outputs.title }}
directly into a JSON string which breaks when titles contain quotes or newlines;
update both occurrences of the JSON --data payload (the "text":"❓ *Question
about <https://github.com/${{ github.repository }}/issues/${{
steps.issue.outputs.number }}|#${{ steps.issue.outputs.number }} — ${{
steps.issue.outputs.title }}>*..." string) to properly JSON-escape the title
(and any other interpolated values) before sending—e.g., construct the payload
using jq or another JSON-safe method so the title is escaped and then post to
"$SLACK_WEBHOOK_URL"; apply the same change to the second similar payload later
in the file.
---
Nitpick comments:
In @.github/workflows/amber-issue-handler.yml:
- Around line 85-86: The workflow requires running the CLI command `gh issue
view --comments` but does not guarantee `gh` is authenticated; update the two
"issue-fix ambient-action" steps that call `gh issue view --comments` to ensure
authentication by either 1) adding a preceding auth/setup step (e.g., use an
`actions/setup-gh` or run `gh auth login --with-token` with a provided secret),
or 2) conditionally checking `gh auth status` and falling back to using the REST
API with `GITHUB_TOKEN`/`secrets.GH_TOKEN`; ensure the `GH_TOKEN` (or
`GITHUB_TOKEN`) secret is passed into the job environment and referenced when
authenticating so the `gh` invocation cannot fail due to missing session auth.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: f119ef0a-2bb5-495d-959f-309ecddc9c49
📒 Files selected for processing (1)
.github/workflows/amber-issue-handler.yml
- Resolve merge conflicts with main branch - Fix JSON injection vulnerability: use jq to safely escape titles - Add GITHUB_TOKEN to environment variables for gh CLI auth - Include PR/issue titles in all Slack notifications - Add title capture to context resolution step Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
I've resolved the merge conflicts and fixed the security issues identified by CodeRabbit: Fixed Issues
Changes ReadyAll fixes are committed locally:
Permission Issue
The local repository is at 🤖 Session |
|
✅ Changes pushed successfully Commit
CI is now running. Waiting for checks to complete. 🤖 Session |
✅ PR Maintenance CompleteAll issues have been resolved and CI is green: Fixed Issues
CI Status
PR State
Next Steps: This PR is ready for final human review and merge. 🤖 Session |
There was a problem hiding this comment.
Actionable comments posted: 1
♻️ Duplicate comments (1)
.github/workflows/amber-issue-handler.yml (1)
700-706:⚠️ Potential issue | 🟠 MajorBatch need-help notifications still use the old payload template.
This is now the last path still posting raw
--data '{"text":...}'and onlyPR #{number}. That misses the title requirement for scheduled PR fixes, and the free-form "what you tried" text can still break JSON here. The loop already haspr["title"]on Lines 636-639, so this prompt should use the samejq -ncpattern as the other Slack templates.Suggested shape
for pr in prs: number = pr["number"] body = pr.get("body", "") + title = pr.get("title", "") fm = parse_frontmatter(body) ... - curl -X POST -H 'Content-type: application/json' \\ - --data '{{"text":"🚨 *Need help with <https://github.com/{REPO}/pull/{number}|PR #{number}>*\\n<reason — what you tried and why you are stuck>\\n\\n<'"$PLATFORM_HOST/projects/$AGENTIC_SESSION_NAMESPACE/sessions/$AGENTIC_SESSION_NAME"'|View Session>"}}' \\ - "$SLACK_WEBHOOK_URL" + PAYLOAD=$(jq -nc --arg text "🚨 *Need help with <https://github.com/{REPO}/pull/{number}|PR #{number} — {title}>* + <reason — what you tried and why you're stuck> + + <$PLATFORM_HOST/projects/$AGENTIC_SESSION_NAMESPACE/sessions/$AGENTIC_SESSION_NAME|View Session>" '{{text: $text}}') + curl -X POST -H 'Content-type: application/json' --data "$PAYLOAD" "$SLACK_WEBHOOK_URL"🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.github/workflows/amber-issue-handler.yml around lines 700 - 706, The batch Slack notification is still using a raw --data JSON string that omits pr["title"] and can break on unescaped free-form text; replace the curl --data '{{"text":"..."} }' usage in the notification block with the same safe jq -nc payload construction used earlier (the pattern that references pr["title"] at lines ~636-639) so the message includes PR title and uses jq -R/--arg to safely embed pr["title"], pr["number"], the user's free-form reason, and the session link variables (PLATFORM_HOST/AGENTIC_SESSION_NAMESPACE/AGENTIC_SESSION_NAME) before piping to curl "$SLACK_WEBHOOK_URL".
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In @.github/workflows/amber-issue-handler.yml:
- Around line 131-132: Remove the full write-scoped GITHUB_TOKEN from the
session environment injection: stop passing "GITHUB_TOKEN" via the
environment-variables payload and instead either fetch required GitHub data on
the runner before starting the model sessions (using the job's token locally) or
supply a separate read-only token scoped to the minimal "issues/PR read" scopes;
update the environment-variables entries (the "environment-variables" key and
its JSON) to exclude "GITHUB_TOKEN" and apply the same change to the other
occurrences mentioned (lines corresponding to the other environment-variables
blocks).
---
Duplicate comments:
In @.github/workflows/amber-issue-handler.yml:
- Around line 700-706: The batch Slack notification is still using a raw --data
JSON string that omits pr["title"] and can break on unescaped free-form text;
replace the curl --data '{{"text":"..."} }' usage in the notification block with
the same safe jq -nc payload construction used earlier (the pattern that
references pr["title"] at lines ~636-639) so the message includes PR title and
uses jq -R/--arg to safely embed pr["title"], pr["number"], the user's free-form
reason, and the session link variables
(PLATFORM_HOST/AGENTIC_SESSION_NAMESPACE/AGENTIC_SESSION_NAME) before piping to
curl "$SLACK_WEBHOOK_URL".
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: 0d14b83b-2d22-4368-9b91-0a0a6808604e
📒 Files selected for processing (1)
.github/workflows/amber-issue-handler.yml
| environment-variables: >- | ||
| {"SLACK_WEBHOOK_URL": "${{ secrets.SLACK_WEBHOOK_URL }}", "PLATFORM_HOST": "${{ secrets.PLATFORM_HOST }}"} | ||
| {"SLACK_WEBHOOK_URL": "${{ secrets.SLACK_WEBHOOK_URL }}", "PLATFORM_HOST": "${{ secrets.PLATFORM_HOST }}", "GITHUB_TOKEN": "${{ secrets.GITHUB_TOKEN }}"} |
There was a problem hiding this comment.
Don't give the session the write-scoped workflow token.
These env vars inject the job's GITHUB_TOKEN into model sessions; in the fix/issue flows, those same sessions are explicitly told to consume untrusted issue/PR text first. That makes prompt injection materially worse, even though the new gh issue/pr view ... / gh issue view --comments behavior only needs reads. Prefer fetching the needed GitHub data on the runner, or pass a separate read-only token scoped just to those reads.
As per coding guidelines, "Verify secrets are not exposed and permissions are scoped."
Also applies to: 352-353, 377-378
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In @.github/workflows/amber-issue-handler.yml around lines 131 - 132, Remove the
full write-scoped GITHUB_TOKEN from the session environment injection: stop
passing "GITHUB_TOKEN" via the environment-variables payload and instead either
fetch required GitHub data on the runner before starting the model sessions
(using the job's token locally) or supply a separate read-only token scoped to
the minimal "issues/PR read" scopes; update the environment-variables entries
(the "environment-variables" key and its JSON) to exclude "GITHUB_TOKEN" and
apply the same change to the other occurrences mentioned (lines corresponding to
the other environment-variables blocks).
Address CodeRabbit review: the batch PR fixer's Slack notification template was the last remaining instance using raw --data JSON, which is vulnerable to JSON injection from unescaped text and omits the PR title. Switched to jq -nc pattern consistent with all other notification templates. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Addressing CodeRabbit Review CommentsBatch Slack notification (lines 700-706) — Fixed ✅Commit d1267bb updates the batch notification template to use GITHUB_TOKEN in environment-variables (lines 131-132, 352-353, 377-378) — Acknowledged, kept as-isThe The token is the workflow's automatic GH_TOKEN for session auth (nitpick, lines 85-86) — Already coveredThe sessions receive 🤖 Session |
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
.github/workflows/amber-issue-handler.yml (1)
72-72: 🛠️ Refactor suggestion | 🟠 MajorPin action to full commit SHA.
ambient-code/ambient-action@v0.0.5uses a tag. Per coding guidelines, pin to the commit SHA for reproducibility and supply-chain security:- uses: ambient-code/ambient-action@v0.0.5 + uses: ambient-code/ambient-action@<full-40-char-sha>Also applies to lines 232, 292, and 361.
As per coding guidelines: "Pin action versions to SHA."
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.github/workflows/amber-issue-handler.yml at line 72, Replace the tagged action reference uses: ambient-code/ambient-action@v0.0.5 with the action pinned to its full commit SHA; search for all occurrences of the string "uses: ambient-code/ambient-action@v0.0.5" (there are multiple instances) and update each to use the corresponding full commit SHA (e.g., ambient-code/ambient-action@<commit-sha>) to ensure reproducible, secure workflow runs.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In @.github/workflows/amber-issue-handler.yml:
- Around line 196-197: The current raw echo into $GITHUB_OUTPUT for the TITLE
variable can break when the PR title contains newlines or percent signs; change
the write to use the GitHub Actions multiline delimiter syntax instead: replace
the echo "title=$TITLE" >> $GITHUB_OUTPUT with the heredoc-style pattern (echo
"title<<EOF" >> $GITHUB_OUTPUT; echo "$TITLE" >> $GITHUB_OUTPUT; echo "EOF" >>
$GITHUB_OUTPUT) so the full value (including newlines and special chars) is
preserved; apply the same fix to the other echo lines that write to
$GITHUB_OUTPUT (the other occurrences that currently use echo to write
variables) so all outputs use the delimiter syntax.
---
Outside diff comments:
In @.github/workflows/amber-issue-handler.yml:
- Line 72: Replace the tagged action reference uses:
ambient-code/ambient-action@v0.0.5 with the action pinned to its full commit
SHA; search for all occurrences of the string "uses:
ambient-code/ambient-action@v0.0.5" (there are multiple instances) and update
each to use the corresponding full commit SHA (e.g.,
ambient-code/ambient-action@<commit-sha>) to ensure reproducible, secure
workflow runs.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: d9c3b9a3-9168-49a6-9f75-871778a27ff9
📒 Files selected for processing (1)
.github/workflows/amber-issue-handler.yml
Titles with newlines or percent characters can corrupt step outputs when written with plain echo. Switch to the GitHub Actions heredoc/delimiter syntax for all three title output locations. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
GITHUB_OUTPUT title safety — Fixed ✅Commit 8338b7e switches all three title output locations (lines 53, 197, 211) to use the GitHub Actions heredoc/delimiter syntax, preventing corruption from titles with newlines or percent characters. E2E test failure — Pre-existing flaky testThe 🤖 Session |
Resolve merge conflicts with main branch changes that added Session URL resolution instructions. Combined our Slack notification templates with main's SESSION_URL placeholder pattern. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
ambient-code
bot
added
ambient-code:needs-human
|
AI was unable to resolve after 3 attempts. Needs human attention. Iterations:
The PR is now healthy (no conflicts, CI was green before merge), but the retry_count has reached 3 due to iterative fixes across multiple review cycles. A human should review the final state and approve. 🤖 Session |
Gkrumbach07
added
ambient-code:managed
|
Reverting the circuit breaker — the PR is healthy (no conflicts, CI green, reviews addressed). The previous comment incorrectly treated normal maintenance iterations as failures. Labels restored: 🤖 Session |
1 participant
Summary
Improve Slack notification formatting across the Amber issue handler workflow:
jq -ncfor JSON-safe payload construction in all Slack notification templates (prevents JSON injection from issue/PR titles containing quotes or newlines)gh issue/pr viewbefore sending notificationsTest plan
check-yamlpre-commit hook)jq -ncpattern consistently🤖 Ambient Session