coldpass

Atrayee-dev/secure-ai-agent-boundary

GitHub ↗ · How this check works

Per-setup public labels: Linux / Haiku: Didn’t work

Static status badge · SVG · read-only JSON. Embed with ![Coldpass status](https://coldpass.dev/badge/Atrayee-dev--secure-ai-agent-boundary.svg).

agent-harness

Didn’t work

The recorded operation failed on 1 setup.

Answers apply to each recorded setup, revision and operation. Earlier observations remain in Run details.

Tested on Linux
PlatformHaikuLuna
Linuxno

Linux / Haiku 5.5: no · Oct 26, 2026 · 410c962288e3 · claude-code / 2.1.310 (Claude Code)
Recorded operation: Pinned snapshot Atrayee-dev/secure-ai-agent-boundary at 410c962288e34d0d3baacc34012cda4af51f0ac9, Linux VM, Python 3.12.3 available. No browser or credentials were needed.

MCP
No
Local or hosted
local
Local: runs on your machine
Maintained
Maintained
actively maintained: last commit 0 days ago; 811 commits and 0 merged PRs in 90 days; 0 open issues, 0 open PRs; no releases

Last commit Oct 26, 2026

Issue
yes, not fixed yet
Our test was blocked; no repo verdict.
Evidence
not graded: our checker rejected this run
The test agent reported: Install and run CodeBoundary as documented in README.md, then exercise one real product operation
Agent interfaces
Not recorded
Not recorded

Observed interfaces were exercised by the checker. Documentation alone does not confirm compatibility.

Tested with
Claude Code / Haiku 5.5
  • Linux: Claude Code / Haiku 5.5 · agent 2.1.310 (Claude Code) · Recorded operation failed · Oct 26, 2026 on 410c962288e3

Each entry describes its recorded operation and source revision.

Safety
No unexpected network or file activity found in these checks

No problems traced to this repository code in these checks; a short run cannot prove safety.

Full event, path and process lists for each observed setup:

Repair history and test corrections
Current diagnosis: unknown · resolved

superseded; repo Atrayee-dev/secure-ai-agent-boundary has a later non-harness result a recorded run

Repair record fix-cb3bf27ad1b428f676e5 · current status JSON. Original run outcomes remain unchanged.

Failed run: Atrayee-dev/secure-ai-agent-boundary · Oct 16, 2026 · Our test failed; no repo verdict.

Repair development: Atrayee-dev/secure-ai-agent-boundary · Oct 16, 2026 · Repair work (development) · not an initial-trial result.

Test-repair rerun: Atrayee-dev/secure-ai-agent-boundary · Oct 17, 2026 · Unsupported in this smoke.

  • Evidence unavailable in the public snapshot.
Run details

Haiku 5.5

Native agent · Recorded operation failed

Operation / attempted scope Pinned snapshot Atrayee-dev/secure-ai-agent-boundary at 410c962288e34d0d3baacc34012cda4af51f0ac9, Linux VM, Python 3.12.3 available. No browser or credentials were needed.

Agent-reported scope: Install and run CodeBoundary as documented in README.md, then exercise one real product operation

Run details
run
Atrayee-dev--secure-ai-agent-boundary-claude-haiku-5-5-1793058313793756339
date
Oct 26, 2026
time to working
not recorded
product hosting / execution
Not recorded
recorded operation scope
Pinned snapshot Atrayee-dev/secure-ai-agent-boundary at 410c962288e34d0d3baacc34012cda4af51f0ac9, Linux VM, Python 3.12.3 available. No browser or credentials were needed.
model
claude-haiku-5-5
agent / version
claude-code / 2.1.310 (Claude Code)
platform
linux
reported cost
$0.008667
Recorded source
https://github.com/Atrayee-dev/secure-ai-agent-boundary
Backend
hyperv
Tool transport
mcp-assigned-vm
Requested source pin (not an observed revision)
{"commit": "410c962288e34d0d3baacc34012cda4af51f0ac9", "url": "https://github.com/Atrayee-dev/secure-ai-agent-boundary"}
Test version
readme-ca496ea10bb07418d1d8f1c256c2433a55547acc0fb62d8b6d86ae33c976fbfa
revision
410c962288e34d0d3baacc34012cda4af51f0ac9
Current diagnosis: unknown · pending

The pinned snapshot contains documentation and a static page only, with no runnable CodeBoundary program.

Repair record fix-95e5d2a301c251497174 · current status JSON. Original run outcomes remain unchanged.

Failed run: Atrayee-dev/secure-ai-agent-boundary · Oct 26, 2026 · Recorded operation failed.

Fix: fix-95e5d2a301c251497174.

Recorded blocker / warning

Step: Locating an installable product: README.md has no runnable command, and no source or package exists in the snapshot.

The pinned snapshot contains documentation and a static page only, with no runnable CodeBoundary program.

Safety: No problems traced to this repository code in these checks.

No problems traced to this repository code in these checks.

OSV scanned supported lockfiles; unpinned dependencies and unsupported formats are outside this check.

A short run cannot establish that code is safe.

Connections show IP addresses; proxies can hide the final hostname.

Tracing covers dispatched commands and their children, not the whole OS; guest root can interfere with tracing.

Full safety evidence →

Answers recorded for this run
  • Install-free operation: The snapshot contains no installable package, source code, entry point, or install command. README.md, index.html, two SVG files and a scheduled GitHub Actions workflow are the only files present.

    Source: CodeBoundary does not require installing a daemon, patching your kernel, or running a heavyweight control plane.

    Observed: The snapshot contains no installable package, source code, entry point, or install command. README.md, index.html, two SVG files and a scheduled GitHub Actions workflow are the only files present.

  • Primary entry point is a config file: No .codeboundary.yml exists in the snapshot and no program in the snapshot reads one, so there is nothing to run or read back.

    Source: The primary entry point is a file called `.codeboundary.yml` at the root of your repository.

    Observed: No .codeboundary.yml exists in the snapshot and no program in the snapshot reads one, so there is nothing to run or read back.

Examples

  • Run the documented CodeBoundary quickstart: README.md has no install, build, run or CLI command section. The only executable-looking material is the GitHub Actions workflow, which is a CI schedule and not product usage.

    Source: README.md

    Expected: documented command completes successfully

    Observed: README.md has no install, build, run or CLI command section. The only executable-looking material is the GitHub Actions workflow, which is a CI schedule and not product usage.

  • Run the repository's own test suite: No test files, package manifest, Makefile or test command exist in the snapshot; nothing was run.

    Source: README.md

    Expected: documented command completes successfully

    Observed: No test files, package manifest, Makefile or test command exist in the snapshot; nothing was run.

Local or hosted: local

Suggested next step (agent-reported): Publish the CodeBoundary source or package in the repository (for example a pyproject.toml or package.json, plus the CLI entry point), add install and quickstart commands to README.md, and include a test or example that runs one real operation such as creating a .codeboundary.yml and reading it back.

What the agent did

Install and run CodeBoundary as documented in README.md, then exercise one real product operation

cd <guest-path> && pwd && ls -la && id -u && python3 --version && (git -C . rev-parse 2>/dev/null || echo "not git") && ls /inputs 2>/dev/null; ls <guest-path> 2>/dev/null
cd <guest-path> && cat README.md 2>/dev/null | head -c 7000; echo; echo ---; find . -maxdepth 2 -not -path './.git*' | head -100
cd <guest-path> && tail -c +7000 README.md | head -c 4500; echo; echo ---; ls -la .github .github/* ; grep -n -i "install\|pip\|npm\|python\|run\|usage\|\$ " README.md | head -40
cd <guest-path> && cat .github/workflows/UvWBoLBblbqu.yml; echo ---; file .github/feXHEbontYwhhU; head -c 300 .github/feXHEbontYwhhU | cat -v | head -20
cd <guest-path> && grep -o -i -E "(pip install|npm |npx |curl [^<\"]*|git clone[^<\"]*|python[0-9]* [^<\"]*|codeboundary [a-z-]*|docker [a-z]*)" index.html | sort | uniq -c | head -30; echo "---title:"; grep -o "<title>[^<]*" index.html; echo "--- find code files:"; find . -type f \( -name "*.py" -o -name "*.js" -o -name "*.ts" -o -name "package.json" -o -name "pyproject.toml" -o -name "setup.py" -o -name "*.yml" -o -name "Makefile" \) | grep -v "^./.coldpass"; echo "--- pages:"; grep -o -i "href=\"[^\"]*\"" index.html | head -15
cd <guest-path> && ls -la .coldpass; grep -n -E "argparse|add_argument|def main|--validate|--replay" <guest-path> | head -30; echo; sed -n 1,30p index.html | head -c 1200
mkdir -p <guest-path> && cat > <guest-path> <<'EOF'
[
  {"kind":"claim","claim":"Install-free operation","source":"CodeBoundary does not require installing a daemon, patching your kernel, or running a heavyweight control plane.","testable":false,"reason":"The snapshot contains no installable package, source code, entry point, or install command. README.md, index.html, two SVG files and a scheduled GitHub Actions workflow are the only files present."},
  {"kind":"claim","claim":"Primary entry point is a config file","source":"The primary entry point is a file called `.codeboundary.yml` at the root of your repository.","testable":false,"reason":"No .codeboundary.yml exists in the snapshot and no program in the snapshot reads one, so there is nothing to run or read back."}
]
EOF
python3 -I -c 'import json,sys; json.load(open(sys.argv[1])); print("claims ok")' <guest-path>
cd <guest-path> && cat > probe.json <<'EOF'
{"kind":"cli","argv":["codeboundary"],"cwd":"<guest-path>","checks":[]}
EOF
cat > cycles.json <<'EOF'
{"detected":false,"reason":"Inspected README.md, index.html and .github/workflows/UvWBoLBblbqu.yml. The only schedule is a GitHub Actions cron ('0 * * * *') that appends a timestamp line to .github/feXHEbontYwhhU and commits it to main; it is a CI schedule, not a product cycle, and it was not run. No product scheduler, daemon, loop or --once entry point exists in the snapshot.","checks":[]}
EOF
cat > examples.json <<'EOF'
[
  {"kind":"example","claim":"Run the documented CodeBoundary quickstart","source":"README.md","reason":"README.md has no install, build, run or CLI command section. The only executable-looking material is the GitHub Actions workflow, which is a CI schedule and not product usage."},
  {"kind":"example","claim":"Run the repository's own test suite","source":"README.md","reason":"No test files, package manifest, Makefile or test command exist in the snapshot; nothing was run."}
]
EOF
python3 -I -c 'import json; [json.load(open(f)) for f in ["probe.json","cycles.json","examples.json"]]; print("ok")' 2>&1; ls -la
cd <guest-path> && cat > report.json <<'EOF'
{"outcome":"blocked","fault_owner":"repo","operation":"Install and run CodeBoundary as documented in README.md, then exercise one real product operation","operation_scope":"Pinned snapshot Atrayee-dev/secure-ai-agent-boundary at 410c962288e34d0d3baacc34012cda4af51f0ac9, Linux VM, Python 3.12.3 available. No browser or credentials were needed.","steps":["Read README.md in full (concept and configuration description only; no install, build, run or CLI commands).","Listed the snapshot: README.md, index.html, button.svg, preview.svg, .github/workflows/UvWBoLBblbqu.yml and .github/feXHEbontYwhhU (a timestamp log file).","Searched index.html and README.md for install/run commands (pip, npm, npx, curl, git clone, python, docker): none found.","Searched for source or package files (*.py, *.js, *.ts, package.json, pyproject.toml, setup.py, Makefile): none found, apart from the workflow file."],"deviations":["No product install was possible: the snapshot has no installer, package or source code, so the README's installer step does not exist.","probe.json uses argv ['codeboundary'] as a placeholder for a nonexistent executable, with checks [] as the unavailable-executable form. No product operation was run, so no product output was observed.","The GitHub Actions workflow (hourly cron committing to main) was inspected but not run, because it is a CI schedule and not a product entry point.","Did not hand-write product code or substitute a made-up CLI."],"stuck_step":"Locating an installable product: README.md has no runnable command, and no source or package exists in the snapshot.","main_blocker":"The pinned snapshot contains documentation and a static page only, with no runnable CodeBoundary program.","plain_failure":"This repository does not contain the CodeBoundary program. The README describes how the tool is meant to work, but it gives no way to install or start it, and the files here are only a description, a web page and an automated timestamp bot. Nothing could be run, so nothing was confirmed to work.","suggested_fix":"Publish the CodeBoundary source or package in the repository (for example a pyproject.toml or package.json, plus the CLI entry point), add install and quickstart commands to README.md, and include a test or example that runs one real operation such as creating a .codeboundary.yml and reading it back."}
EOF
python3 -I -c 'import json; json.load(open("report.json")); print("report ok")'; cd <guest-path> && <env-var>=<guest-path> python3 -I <guest-path> --validate --replay; echo "exit=$?"

No product verdict. Run record.

No product verdict. Run record.

No product verdict. Run record.

No product verdict. Run record.

No product verdict. Run record.

No product verdict. Run record.

No product verdict. Run record.