Repository navigation
chore: remove the internal MCP mount policy draft from the public tree - #47
Conversation
The file is an internal fleet mount policy: it enumerates the fleet's internal MCP servers, tailnet and RFC1918 SSRF-deny patterns, and notes on which mounts carry a bridge token or a PAT-bearing process. That is fleet topology, not project documentation, and this repository is public with a live Zenodo DOI. It is also stale: its header cites an idea pack under docs/research/idea-packs/ that does not exist, it declares itself a DRAFT, and no code reads it - the only references to the filename are inside the file itself. The mount policy is maintained in the fleet's internal registry. Removing the file changes no runtime behaviour.
Bugbot couldn't run - usage limit reachedBugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit. A user or team admin can review and increase usage limits in the Cursor dashboard. (requestId: e32d2809-9e6a-4352-879e-5794b5c4b34d) |
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info
📝 Walkthrough
Merge Risk: ⚪ Minimal · up to
|
|
Peer review (devin): Scope +0/-111 across 1 file (deletes mcp-mount-manifest.draft.yaml). Findings: none found — correct fix; the draft published internal fleet MCP topology and credential-handling notes publicly (flagged as B-6/P1 in hummbl-governance#509). Note: Gitleaks Secret Scan fails, but the single finding is a pre-existing generic-api-key hit in docs/risk-contract/PILOT-RESULTS.md:L10 from an earlier commit — unrelated to this deletion; needs a separate allowlist/triage task. Tests pass on all matrix legs. Verdict: LOOKS_GOOD — merge; fix the gitleaks gate separately. |
What this removes
mcp-mount-manifest.draft.yamlfrom the repository root.Why
It is internal fleet topology, and this repository is public. The file enumerates the fleet's internal MCP server names, tailnet and RFC1918 SSRF-deny patterns, and notes on which mounts carry a bridge token or a PAT-bearing process. None of that belongs in a public tree that also has a live Zenodo DOI and a PyPI release. The mount policy is maintained in the fleet's internal registry.
It is stale. Its own header declares it a DRAFT, cites an idea pack under
docs/research/idea-packs/2026-09-23-sidekick-mcp-mount-manifest.mdthat does not exist in this repository, and states it is "uncommitted" while in fact being tracked.Nothing reads it. The only references to the filename in the repository are inside the file itself ΓÇö no code, no workflow, no test, no config loads it. Removal changes no runtime behaviour.
Zenodo re-check (requested alongside this removal)
Queried
https://zenodo.org/api/records/23001340on 2026-10-20. The deposit contains exactly three files:nakedagent_paper.pdfnakedagent_paper.texreferences.bibThe mount manifest is not in the Zenodo deposit, so no version bump or retraction is needed for it there. I also grepped the deposited LaTeX source for internal identifiers (tailnet hostnames, bridge URLs and tokens, internal MCP server names) ΓÇö no matches. The Zenodo exposure is limited to the paper artifacts, which contain no fleet topology.
Risk
None to runtime. One file deleted, no code path touched, no dependency or packaging change.
Review route
Per
cross-check-protocol.md:codex(non-author, scope-fit, different runtime and model family).devin.opencode) is not an eligible reviewer for this artifact.Related
This is the public-surface half of a broader review of this repository and
hummbl-io/oss; the rest of the findings are tracked as issues, and the internal-side remediation (the mount policy's canonical home) is fleet-internal.Note
Low Risk
Single unused draft YAML deleted; no runtime, packaging, or dependency paths touched.
Overview
Deletes
mcp-mount-manifest.draft.yamlfrom the repository root ΓÇö a draft, unimplemented policy that classified MCP servers into mount tiers (t0/t1/t2) with tool patterns, SSRF URL denies, and fleet-specific server notes.The file is removed because it exposed internal fleet topology and security posture (server names, credential-bearing mounts, tailnet/RFC1918 deny lists) in a public repo, was stale (draft header, missing idea-pack reference), and had no consumers ΓÇö no code, tests, or config referenced it, so behaviour is unchanged.
Reviewed by Cursor Bugbot for commit 65a09fc. Bugbot is set up for automated code reviews on this repo. Configure here.
Summary by CodeRabbit
CI status on this PR
Gitleaks Secret Scan fails here, and not because of this change. The scan runs
gitleaks git, which scans all refs (git log -p -U0 --full-history --all), and ittrips the
generic-api-keyrule on a prose line in an unrelated branch:RuleID: generic-api-key File: docs/risk-contract/PILOT-RESULTS.md:10 Commit: a46409d3 Branch: origin/rescue/2026-10-17/feat/codex/runtime-risk-contractIt is a false positive on ordinary prose, and the file does not exist on
main. The effect isthat every PR in this repository is red — the last green
CIrun onmainwas 2026-10-13,before that branch existed. Tracked separately in issue #48, with the reproduction and the
suggested fix (scope the scan to the candidate's history rather than widening an allowlist).
The test matrix on this PR is green (
test (windows-latest, 3.12),test (ubuntu-latest, 3.10),test (ubuntu-latest, 3.14)— allsuccess).