Skip to content

chore(github): Change range of Write Tool#21743

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
s1gr1d merged 1 commit into from
Jun 24, 2026
Merged

chore(github): Change range of Write Tool #21743

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .github/workflows/track-framework-updates.yml
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -60,7 +60,7 @@
claude_args: |
--model claude-opus-4-8
--max-turns 80
--allowedTools "Read,Write(/.agents/skills/track-framework-updates/output/**),Bash(python3 .agents/skills/track-framework-updates/scripts/collect_updates.py *),Bash(python3 .agents/skills/track-framework-updates/scripts/check_support.py)"
--allowedTools "Read,Write,Bash(python3 .agents/skills/track-framework-updates/scripts/collect_updates.py *),Bash(python3 .agents/skills/track-framework-updates/scripts/check_support.py)"

Check warning on line 63 in .github/workflows/track-framework-updates.yml

View check run for this annotation

@sentry/warden/ warden: security-review

Claude agent given unrestricted workspace Write access, enabling prompt-injection-driven script overwrite 

Removing the `Write(/.agents/skills/track-framework-updates/output/**)` path restriction gives the Claude agent unrestricted write access to the entire workspace; an adversary who injects instructions via GitHub issue/PR content that Claude reads could overwrite `.agents/skills/track-framework-updates/scripts/write_job_summary.py`, which the next step then executes unconditionally with `python3 "$SCRIPT"`, achieving runner-level code execution and potential exfiltration of `ANTHROPIC_API_KEY` and the `id-token: write`-scoped `GITHUB_TOKEN`.

@sentrysentryBotJun 24, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Bug: Removing the path restriction on the Write tool for the Claude agent relies solely on prompt-based guardrails, allowing it to potentially overwrite files and cause workflow failures.
Severity: LOW

Suggested Fix

Re-introduce a technical restriction on the Write tool if possible, or add more robust checks to validate the agent's output. If the unrestricted Write is necessary, ensure the prompt is as clear as possible and consider adding a step to verify that no unintended files were modified before proceeding.

Prompt for AI Agent
Review the code at the location below. A potential bug has been identified by an AI
agent. Verify if this is a real issue. If it is, propose a fix; if not, explain why it's
not valid.
Location: .github/workflows/track-framework-updates.yml#L64
Potential issue: The workflow configuration was changed to give the Claude AI agent's
`Write` tool unrestricted access, removing the previous path restriction
`Write(/.agents/skills/track-framework-updates/output/**)`. This replaces a technical
control with a prompt-based instruction, which is a weaker guardrail. While the agent is
instructed to only write to a specific output directory, a misinterpretation or model
deviation could lead it to overwrite or corrupt other files in the ephemeral workspace,
such as scripts or configuration files. This could cause the workflow to fail. The
impact is limited to the single workflow run as the environment is temporary.

Did we get this right? 👍 / 👎 to inform future reviews.

- name: Post job summary
if: always()
Expand Down
Loading