sentinelprobe 0.9.0__tar.gz

sentinelprobe-0.9.0/LICENSE

@@ -0,0 +1,22 @@
+MIT License
+
+Copyright (c) 2026 SentinelProbe contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+

sentinelprobe-0.9.0/MANIFEST.in

@@ -0,0 +1 @@
+include docs/*.md

sentinelprobe-0.9.0/PKG-INFO

@@ -0,0 +1,147 @@
+Metadata-Version: 2.4
+Name: sentinelprobe
+Version: 0.9.0
+Summary: Authorized AI prompt-injection and agent-boundary testing CLI
+Author: SentinelProbe contributors
+License-Expression: MIT
+Project-URL: Homepage, https://github.com/jashidsany/sentinelprobe
+Project-URL: Repository, https://github.com/jashidsany/sentinelprobe
+Project-URL: Issues, https://github.com/jashidsany/sentinelprobe/issues
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: PyYAML>=6.0.1
+Provides-Extra: browser
+Requires-Dist: playwright>=1.44.0; extra == "browser"
+Dynamic: license-file
+
+# SentinelProbe
+
+SentinelProbe is a CLI for authorized AI prompt-injection and agent-boundary testing. It runs repeatable YAML cases against AI assistants, enterprise search tools, browser-based AI apps, HTTP test shims, and local CLI agents, then scores responses with deterministic checks.
+
+Use it only with approved systems, approved accounts, fake documents, and fake secrets.
+
+## Install
+
+From source:
+
+```bash
+python3 -m pip install .
+```
+
+For isolated installs:
+
+```bash
+pipx install .
+```
+
+For browser automation:
+
+```bash
+python3 -m pip install '.[browser]'
+python3 -m playwright install chromium
+```
+
+After PyPI publishing:
+
+```bash
+pipx install sentinelprobe
+```
+
+## Quick Start
+
+Check local setup:
+
+```bash
+sentinelprobe doctor
+sentinelprobe doctor --target claude-code
+sentinelprobe doctor --target browser
+```
+
+List bundled suites:
+
+```bash
+sentinelprobe list-suites
+```
+
+Run the local mock baseline:
+
+```bash
+sentinelprobe run --cases builtin --provider mock --verbose
+```
+
+Run Claude Code with response-only defaults:
+
+```bash
+sentinelprobe claude-code
+```
+
+Run file-based coding-agent prompt injection against Claude Code:
+
+```bash
+sentinelprobe claude-code --test agent-files --agent-files --verbose --only-findings --html-report
+```
+
+Limit cost during smoke tests:
+
+```bash
+sentinelprobe claude-code --test indirect --mutations --limit 5 --verbose --only-findings
+```
+
+Create HTML and trace artifacts:
+
+```bash
+sentinelprobe claude-code --test agent-files --agent-files --html-report --trace-file reports/agent_files_trace.txt
+```
+
+Compare two reports:
+
+```bash
+sentinelprobe compare --before reports/baseline.json --after reports/latest.json --html-report
+```
+
+Run source-tree regression checks during development:
+
+```bash
+python3 scripts/check.py
+python3 scripts/check.py --build --wheel-smoke
+```
+
+## Test Suites
+
+- `direct-basic`: basic direct prompt injection cases.
+- `direct-advanced`: advanced direct prompt injection cases.
+- `direct`: basic plus advanced direct prompt injection.
+- `indirect`: inline retrieved-content prompt injection cases.
+- `agent-files`: file-based coding-agent prompt injection cases.
+- `builtin`: all bundled cases.
+
+Use `--mutations` to expand suites that define deterministic variants.
+
+## Providers
+
+- `mock`: local safe baseline.
+- `http`: approved API or internal test shim.
+- `command`: local wrapper around a CLI agent.
+- `browser`: Playwright-driven browser session for approved browser-based AI tools.
+
+## Reports
+
+SentinelProbe writes JSON reports by default. Add `--html-report` for a portable review artifact and `--trace-file` for full prompt and response evidence.
+
+Findings use:
+
+- `pass`: no deterministic issue found.
+- `review`: suspicious output or incomplete safe handling that needs human triage.
+- `fail`: deterministic unsafe behavior, including fake secret leakage.
+
+## Documentation
+
+Detailed usage, provider setup, case format, scoring behavior, browser workflow, and PyPI notes are in [docs/usage.md](docs/usage.md).
+
+## Safety Scope
+
+- Use only approved systems and accounts.
+- Use fake documents, fake secrets, and sandbox data.
+- Do not test destructive actions, credential theft, persistence, malware, or bypass logic.
+- Treat automated findings as triage signals that need manual validation.

sentinelprobe-0.9.0/README.md

@@ -0,0 +1,130 @@
+# SentinelProbe
+
+SentinelProbe is a CLI for authorized AI prompt-injection and agent-boundary testing. It runs repeatable YAML cases against AI assistants, enterprise search tools, browser-based AI apps, HTTP test shims, and local CLI agents, then scores responses with deterministic checks.
+
+Use it only with approved systems, approved accounts, fake documents, and fake secrets.
+
+## Install
+
+From source:
+
+```bash
+python3 -m pip install .
+```
+
+For isolated installs:
+
+```bash
+pipx install .
+```
+
+For browser automation:
+
+```bash
+python3 -m pip install '.[browser]'
+python3 -m playwright install chromium
+```
+
+After PyPI publishing:
+
+```bash
+pipx install sentinelprobe
+```
+
+## Quick Start
+
+Check local setup:
+
+```bash
+sentinelprobe doctor
+sentinelprobe doctor --target claude-code
+sentinelprobe doctor --target browser
+```
+
+List bundled suites:
+
+```bash
+sentinelprobe list-suites
+```
+
+Run the local mock baseline:
+
+```bash
+sentinelprobe run --cases builtin --provider mock --verbose
+```
+
+Run Claude Code with response-only defaults:
+
+```bash
+sentinelprobe claude-code
+```
+
+Run file-based coding-agent prompt injection against Claude Code:
+
+```bash
+sentinelprobe claude-code --test agent-files --agent-files --verbose --only-findings --html-report
+```
+
+Limit cost during smoke tests:
+
+```bash
+sentinelprobe claude-code --test indirect --mutations --limit 5 --verbose --only-findings
+```
+
+Create HTML and trace artifacts:
+
+```bash
+sentinelprobe claude-code --test agent-files --agent-files --html-report --trace-file reports/agent_files_trace.txt
+```
+
+Compare two reports:
+
+```bash
+sentinelprobe compare --before reports/baseline.json --after reports/latest.json --html-report
+```
+
+Run source-tree regression checks during development:
+
+```bash
+python3 scripts/check.py
+python3 scripts/check.py --build --wheel-smoke
+```
+
+## Test Suites
+
+- `direct-basic`: basic direct prompt injection cases.
+- `direct-advanced`: advanced direct prompt injection cases.
+- `direct`: basic plus advanced direct prompt injection.
+- `indirect`: inline retrieved-content prompt injection cases.
+- `agent-files`: file-based coding-agent prompt injection cases.
+- `builtin`: all bundled cases.
+
+Use `--mutations` to expand suites that define deterministic variants.
+
+## Providers
+
+- `mock`: local safe baseline.
+- `http`: approved API or internal test shim.
+- `command`: local wrapper around a CLI agent.
+- `browser`: Playwright-driven browser session for approved browser-based AI tools.
+
+## Reports
+
+SentinelProbe writes JSON reports by default. Add `--html-report` for a portable review artifact and `--trace-file` for full prompt and response evidence.
+
+Findings use:
+
+- `pass`: no deterministic issue found.
+- `review`: suspicious output or incomplete safe handling that needs human triage.
+- `fail`: deterministic unsafe behavior, including fake secret leakage.
+
+## Documentation
+
+Detailed usage, provider setup, case format, scoring behavior, browser workflow, and PyPI notes are in [docs/usage.md](docs/usage.md).
+
+## Safety Scope
+
+- Use only approved systems and accounts.
+- Use fake documents, fake secrets, and sandbox data.
+- Do not test destructive actions, credential theft, persistence, malware, or bypass logic.
+- Treat automated findings as triage signals that need manual validation.