Repository navigation
fix(public-tree): no setup-mode preamble or internal MCP service in the tree, and a guard - #35
Conversation
…he tree, and a guard Six script and config files carried a preamble naming the internal setup mode; each now says it is owned by this repository's maintainers. .mcp.json no longer names our own MCP service, and the scorecard drops its platform run id. bin/check-public-tree (bash) guards the tree, as a "Public tree" step in the lint job and a bin/check stage: no internal host on our domain, no private repository or bare private tracker reference, no one's home directory, no setup-mode name, and none of the private words, which come from the PUBLIC_TREE_PRIVATE_WORDS CI secret, never from the tree. --self-test proves each rule fires and pins the workflow's exact wiring. Closes #34 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`sed | grep -q` under pipefail fails falsely when grep matches before sed has written the rest of a large workflow file (found on magik's). The pin now greps a here-string, like every other check. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Fix failing CI on #35
Queued — waiting for a free box. 🤖 developerz.ai, an automated agent, not a human — developerz-ai[bot] |
…s); guard catches up bin/check's stage list is pinned by CheckTest/CheckScriptTest, so the guard stays a step of the lint job only, which runs on every PR. The guard gains what review on magik#32 found: raw.githubusercontent.com and api.github.com/repos references, tracked paths checked like content, and fix: lines on every failure. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
✅ Reviewed — nothing blocking0 actionable comment(s) · grounded on your code 🤖 developerz.ai — automated review, running on your model and your box. |
There was a problem hiding this comment.
Review summary — 9 file(s), 1 finding(s).
Critical 0 · Major 0 · Minor 1 · Nit 0
Removes setup-mode preambles and the internal MCP server from tracked files, and adds a bash CI guard (with a self-test) that scans tracked files for private hosts, repos, home dirs and secret-supplied words.
1 finding(s) outside this pull request's change (reported, not holding the merge)
.env.example:4minor · docs This PR removes thedeveloperzserver from.mcp.json, so this comment is now wrong: nothing in the tree readsDEVELOPERZ_API_KEYany more. Delete the comment and the variable, or reword the comment to say it is unused.
These sit on code this pull request did not change, so they do not hold the merge.
- test files were not shown to the reviewer — this diff touched no path recognised as a test, so no assertion was available to read the change against
🤖 developerz.ai review — automated, running on your model and your box. What is this?
|
CI green, ready to merge. 🤖 Posted by developerz.ai — an automated agent, not a human. |
…e guard Review on #35: with our MCP service gone from .mcp.json, nothing reads DEVELOPERZ_API_KEY, so .env.example no longer lists it. The guard compares resolved paths (`pwd -P`), so a checkout under a symlinked directory is not mistaken for another tree. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Both fixed in |
There was a problem hiding this comment.
Review summary — 2 file(s), 0 finding(s).
This increment drops the dormant DEVELOPERZ_API_KEY from .env.example and makes bin/check-public-tree resolve its repo root with pwd -P, which fixes the symlinked-checkout false failure flagged earlier. Low risk.
- test files were not shown to the reviewer — this diff touched no path recognised as a test, so no assertion was available to read the change against
🤖 developerz.ai review — automated, running on your model and your box. What is this?
|
CI passed, but approval from a reviewer is required before this merge can proceed. Once approved, the platform's merge gate will handle merging. 🤖 Posted by developerz.ai — an automated agent, not a human. |
Closes #34.
What changes
.env.example,.githooks/pre-commit,bin/dev,bin/stop, and two.dz/files): each now says it is owned by this repository's maintainers..mcp.json: our own MCP service removed;codegraphstays. Scorecard: the platform run id line is gone.bin/check-public-tree(new; bash, no dependencies), run as a "Public tree" step in the CI lint job, which runs on every PR, from the repository root. It is not abin/checkstage, becauseCheckTest/CheckScriptTestpinbin/check's exact stage list. It readsgit ls-files, proves it read itself, and fails on an internal host on our domain, a private repository or bare private tracker reference, someone's home directory, the setup mode's name, or a private word. A CODEOWNERS@org/teamis not a repository. A failure names the file and the rule, never the value.raw.githubusercontent.comandapi.github.com/reposreferences as repository references, and prints afix:line on every failure.--self-testproves each rule fires, including on a 200 KB file with the match on line 1 (no SIGPIPE false pass underpipefail). It also pins CI's exact env lines: commenting one out or weakening the condition fails it.PUBLIC_TREE_PRIVATE_WORDSsecret, never from the tree (fragments joined in a public file are the word: db-mcp-gateway#350). The workflow step passes it and setsPUBLIC_TREE_PRIVATE_WORDS_REQUIRED=1on this repository's own runs; forks and Dependabot skip with a notice. The secret is set.Verification
bin/check-public-tree --self-testread 344 of 366 tracked files, 2 private words(the rest are binary), greenmain's tree.mcp.jsonactionlintThe values are deliberately not repeated here.
🤖 Generated with Claude Code