codefence 1.0.0__py3-none-any.whl

codefence-1.0.0.data/data/README.md

@@ -0,0 +1,299 @@
+# CodeFence
+
+**A tiny offline policy gate for code.**
+
+Check AI-generated code before it reaches Git.
+
+---
+
+## What it does
+
+CodeFence is a single-file, zero-dependency Python CLI that scans
+source code for selected dangerous patterns before you commit. It
+runs a curated set of rules and reports findings with:
+
+- severity (critical / high / medium / low / info)
+- exact location (file:line:column)
+- code snippet
+- short remediation
+- typical before/after fix example
+
+It is designed around the risks commonly encountered in AI-assisted
+development, but scans ordinary source code regardless of how it
+was written.
+
+**The gate workflow:**
+
+    AI writes code
+         |
+         v
+    CodeFence runs on staged files
+         |
+         v
+    policy evaluation (allow / warn / block)
+         |
+         v
+    PASS or BLOCKED
+         |
+         v
+    git commit
+
+## What it does NOT do
+
+- Not an autofixer. It shows a typical fix; it does not edit files.
+- Not a security audit. It is a pattern-based sanity checker.
+- Not a replacement for Semgrep, CodeQL, or professional review.
+- Not interprocedural. No dataflow, no taint analysis.
+- Not a guarantee. A clean scan does not mean the code is secure.
+
+## Install
+
+    pip install codefence
+
+Or run it directly:
+
+    python3 codefence.py app.py
+
+## Quick start
+
+**One-time setup in a git repository:**
+
+    codefence init
+
+This installs a pre-commit hook and creates .codefence/config.json.
+Every commit is then gated automatically.
+
+**Daily usage:**
+
+    codefence --staged          # scan only what is staged
+    codefence --staged --diff   # only NEW findings since baseline
+    codefence baseline          # snapshot current findings
+
+**One-off scans:**
+
+    codefence src/
+    codefence --format json --output report.json src/
+    codefence --format html --output report.html src/
+    codefence --format sarif --output results.sarif .
+
+**Policy-as-code:**
+
+    codefence --policy company-policy.json --staged
+    codefence policy validate company-policy.json
+
+**Explanation:**
+
+    codefence explain R002
+## The 30 rules
+
+Rules are shipped in rules.json (human-readable JSON). You can
+inspect them, and you can add or disable rules for your own use.
+
+### Secrets
+- R001 - Hardcoded API keys / secrets
+- R011 - Hardcoded database credentials
+- R028 - Env var with hardcoded fallback secret
+
+### Injection
+- R002 - SQL injection via string concatenation (Python)
+- R003 - eval() / exec() on untrusted input (Python)
+- R004 - Command injection via shell=True (Python)
+- R005 - innerHTML with unsanitized data (JavaScript)
+- R010 - Path traversal
+- R014 - Prototype pollution (JavaScript)
+- R019 - Unsafe deserialization (pickle, yaml.load)
+
+### Authentication
+- R006 - Missing authentication on endpoints
+- R015 - Missing rate limiting on public endpoints
+- R018 - Open redirect
+- R020 - Never-expiring tokens / JWT
+
+### Cryptography
+- R008 - Insecure random for security tokens (Python)
+- R009 - Disabled TLS/SSL verification
+- R016 - MD5 / SHA1 for password hashing
+
+### Configuration
+- R007 - Overly permissive CORS (*)
+- R017 - Debug mode left enabled
+- R021 - Missing security headers (Express)
+
+### Reliability
+- R022 - Race conditions (check-then-act)
+- R026 - HTTP request without timeout
+- R027 - Naive datetime (no timezone, context-aware)
+
+### Quality
+- R012 - Broad except: pass
+- R013 - Mutable default arguments
+- R023 - Logging sensitive data
+- R024 - Unused imports / dead code
+- R025 - Promise rejection ignored (JavaScript)
+
+### AI-specific
+- R030 - Typosquatted package imports
+- R031 - TODO / FIXME / HACK in new code
+
+Each rule carries a confidence level and a stable rule_version +
+fingerprint for baseline stability.
+
+## Configuration
+
+Options can be passed as CLI flags or through a JSON config file
+(auto-discovered at .codefence/config.json after 'codefence init').
+
+Example .codefence/config.json:
+
+    {
+      "schema": "codefence/config-v1",
+      "format": "json",
+      "severity": "medium",
+      "include": ["*.py", "*.js"],
+      "exclude": ["node_modules", ".git", "venv", "samples"],
+      "max_size": 2097152,
+      "cache": false
+    }
+
+Priority: CLI flags > config file > built-in defaults.
+
+### Optional cache
+
+Pass --cache to enable a local cache at ~/.cache/codefence/. The
+cache is keyed by SHA-256 of the file content plus a fingerprint
+of rules.json. It is off by default. See SECURITY.md for how the
+cache is protected.
+
+### Optional history
+
+Pass --history to record this scan in a local SQLite database at
+~/.local/share/codefence/history.db. Off by default.
+
+Set CODEFENCE_NO_HISTORY=1 to disable history entirely, even when
+--history is set.
+
+Query history:
+
+    codefence history
+    codefence stats
+## Security and privacy
+
+- Zero network calls. Zero.
+- Zero telemetry.
+- Zero data upload.
+- Writes only to --output, and optionally to:
+    ~/.cache/codefence/            (only when --cache is set)
+    ~/.local/share/codefence/      (only when --history is set)
+- Never writes to the files you scan.
+- Never executes the code you scan.
+
+The full threat model is documented in SECURITY.md.
+
+## Verifying the download
+
+Every shipped file is hashed in CHECKSUMS.txt. To verify:
+
+    sha256sum codefence.py rules.json
+
+Compare the output with the corresponding lines in CHECKSUMS.txt.
+If the hashes do not match, do not use the file.
+
+## Payment and pricing
+
+**$12 USD, one-time.**
+
+Includes the current major version (v1.x) and its maintenance
+releases. No subscription. No support. No account.
+
+Payments are available in cryptocurrency only:
+
+- USDT (TRC20 or BEP20)
+- USDC
+- BTC
+- TRX
+- XRP
+
+Official purchase channels:
+
+- Getly
+- SilkRoadx402
+- ctlx.cc
+
+If you find CodeFence on any other site claiming to sell it,
+treat that site as unofficial.
+
+## License
+
+Source-available. Not open source.
+
+Summary:
+
+- Personal use on up to 3 devices that you own or control.
+- Read, study, and privately modify the source.
+- No redistribution. No resale. No bundling.
+- No scanning-as-a-service on a commercial basis.
+
+For plain-language answers to common questions, see LICENSE_FAQ.md.
+For the full legal terms, see LICENSE.txt and TERMS_OF_USE.md.
+
+**Governing law:** England and Wales. Mandatory consumer
+protections in your country of residence remain fully applicable.
+
+## Support
+
+There is no support.
+
+- No email. No chat. No issue tracker.
+- No guaranteed updates.
+- No bug-fix commitments.
+
+If you need a product with ongoing support, please look elsewhere.
+
+## FAQ
+
+**Q: Will this find every security issue in my code?**
+A: No. It finds a curated set of dangerous patterns. It is a
+   sanity check, not a security audit.
+
+**Q: Can I use it in CI?**
+A: Yes. JSON and SARIF outputs are pure on stdout. Exit codes are
+   stable.
+
+**Q: Does it phone home?**
+A: No. Zero network calls. You can verify by reading the source
+   or by running it under strace.
+
+**Q: Does it fix my code?**
+A: No. It shows a typical before/after fix. You apply it yourself.
+
+**Q: Does it support TypeScript?**
+A: No. Python, JavaScript (.js, .mjs, .cjs).
+
+**Q: What about false positives?**
+A: Pattern-based scanners always produce some. Rules with lower
+   confidence are flagged as such. You can disable any rule in
+   rules.json or via policy overrides.
+
+**Q: Can I modify the source?**
+A: Yes, privately. See LICENSE_FAQ.md.
+
+**Q: Can my team use it?**
+A: Each developer using CodeFence on their own machine needs
+   their own license. One license covers one individual on up to
+   three devices.
+
+**Q: How do I get updates?**
+A: There are no guaranteed updates.
+
+**Q: Can I get a refund?**
+A: See REFUND_POLICY.md.
+
+## A note on honesty
+
+This product is built to do a specific, limited thing well. It
+does not overstate what it can do. Every claim in this README is
+traceable to a line of code or a test. If you find a claim that
+is not supported by what the tool actually does, please report
+it.
+
+*End of README.*

codefence-1.0.0.data/data/REFUND_POLICY.md

@@ -0,0 +1,131 @@
+# Refund Policy
+
+**CodeFence — v1.0.0**
+Effective 2026-09-16
+
+---
+
+## Plain summary
+
+- Digital product, delivered instantly as a downloadable file.
+- **EU / UK consumers:** 14-day right of withdrawal, **if the file
+  has not been downloaded**.
+- Once downloaded, the withdrawal right is waived, as permitted by
+  EU Consumer Rights Directive 2011/83/EU, Article 16(m).
+- Statutory consumer protections always apply and are never waived.
+
+---
+
+## 1. Right of Withdrawal (EU and UK consumers)
+
+If you are a consumer resident in the European Union or the United
+Kingdom, you have the right to withdraw from your purchase within
+**fourteen (14) days** of the transaction date, **without giving any
+reason**, **provided that you have not downloaded the digital
+content**.
+
+To exercise this right, you must:
+
+1. Contact the platform through which you purchased the Software
+   (Getly, SilkRoadx402, or ctlx.cc) before the 14-day period
+   expires.
+2. State clearly your intention to withdraw.
+3. Not have downloaded the Software file.
+
+The refund will be processed via the same payment method used for
+the purchase, within a reasonable time.
+
+### Waiver after download
+
+By downloading the Software, you expressly agree and acknowledge
+that:
+
+- The digital content has been supplied to you.
+- You lose your right of withdrawal with respect to the downloaded
+  content.
+
+This waiver is permitted under Article 16(m) of Directive 2011/83/EU
+and equivalent UK law (Consumer Contracts Regulations 2013).
+
+**If you have not downloaded the Software and want a refund, do not
+download the file first.**
+
+---
+
+## 2. Conformity Guarantee
+
+If the Software does not conform to its description at the time of
+purchase, you may be entitled to a remedy under applicable EU/UK
+consumer protection law. This includes, where applicable:
+
+- repair or replacement of the non-conforming digital content,
+- a proportionate price reduction,
+- termination of the contract and full refund.
+
+This right is separate from the 14-day withdrawal right and is not
+limited by it.
+
+---
+
+## 3. Non-Refundable Cases (after download)
+
+After the file has been downloaded, refunds are **not provided**
+for:
+
+- Change of mind.
+- The Software did not detect a specific issue you hoped it would
+  detect. (The Software is a pattern-based checker; its scope is
+  documented in `README.md` and `SECURITY.md`.)
+- False positives or false negatives produced by the pattern rules.
+- Incompatibility with your specific environment that is outside the
+  documented supported platforms.
+- The Software's stated limitations (no autofix, no TypeScript, no
+  interprocedural analysis, etc.). These limitations are documented
+  publicly before purchase.
+- Unwillingness to use the Software after reading its documentation.
+
+---
+
+## 4. Fraud and Abuse
+
+Refunds will not be processed if:
+
+- The purchase was made with a fraudulent payment method.
+- The Software was redistributed or resold in violation of
+  `LICENSE.txt`.
+- The refund request is part of a pattern of abuse.
+
+---
+
+## 5. Statutory Rights
+
+Nothing in this policy limits any statutory right or remedy available
+to you under mandatory consumer protection law in your country of
+residence. If any part of this policy conflicts with such mandatory
+law, the mandatory law prevails.
+
+---
+
+## 6. How to Request a Refund
+
+Refund requests must be made through the platform where the purchase
+was made:
+
+- **Getly** — use the platform's refund mechanism.
+- **SilkRoadx402** — use the platform's dispute mechanism.
+- **ctlx.cc** — use the platform's refund mechanism.
+
+There is no direct email or contact channel for refund requests. This
+is intentional and consistent with the product's stated design.
+
+---
+
+## 7. Timeframe
+
+Refund processing times depend on the platform and the payment
+network. Cryptocurrency refunds are typically processed within a few
+business days once approved by the platform.
+
+---
+
+*End of Refund Policy.*

codefence-1.0.0.data/data/SECURITY.md

@@ -0,0 +1,142 @@
+# Security Model — CodeFence
+
+This document describes, honestly and completely, what this tool can
+and cannot do with respect to security. It is written for users and
+security reviewers.
+
+## What this tool is
+
+CodeFence is a **pattern-based sanity checker**. It reads source
+files, runs pre-compiled regular expressions and a bounded AST walk,
+and reports findings. It is a *helper*, not a security guarantee.
+
+## Threat model
+
+### What the tool does NOT do
+
+- It does **not** execute the scanned code.
+- It does **not** import user modules.
+- It does **not** make network connections.
+- It does **not** send telemetry or data anywhere.
+- It does **not** write into user source files.
+- It does **not** create or modify anything outside:
+  - stdout, or the file passed to `--output`
+  - (optionally, if `--cache` is enabled) `~/.cache/codefence/`
+  - (optionally, if `--history` is enabled) `~/.local/share/codefence/`
+
+### Attack surfaces and how they are handled
+
+1. **Malicious `rules.json`**
+   - Schema is validated on load: rule IDs must match `R###`, severity
+     and language values must be from the allowed sets, regex patterns
+     are compiled eagerly and rejected on error.
+   - Per-pattern length is capped (`MAX_REGEX_PATTERN_LEN` = 500).
+   - Per-rule pattern count is capped (`MAX_REGEX_PATTERNS_PER_RULE` = 20).
+   - Every regex call is run line-by-line, with a hard per-call timeout
+     (`REGEX_TIMEOUT_SEC` = 0.5 s) and per-line cap
+     (`MAX_REGEX_LINE_LEN` = 8192 chars). This bounds catastrophic
+     backtracking (ReDoS).
+
+2. **Malicious source file**
+   - File size is capped by `--max-size` (default 2 MiB).
+   - Files are read as text with `errors="replace"`.
+   - Python files are parsed by CPython's own `ast.parse`. The AST
+     parser is memory-safe in supported Python versions.
+   - JavaScript files are tokenized by our own lexer, written in pure
+     Python. It never executes input.
+
+3. **Cache poisoning (only when `--cache` is used)**
+   - Cache directory is created with mode `0o700`.
+   - Cache files are created with mode `0o600` via `mkstemp` +
+     `os.replace` (atomic).
+   - Cache entries are keyed by SHA-256 of (version, language, rules
+     fingerprint, file content). A modified file produces a different
+     key.
+   - On load, the cache file is checked: must be a regular file (no
+     symlinks), JSON must parse, and each finding must validate. Any
+     failure results in a silent cache miss and a fresh scan.
+   - Cache never contains executable content; it is plain JSON.
+
+4. **History database (only when `--history` is used)**
+   - SQLite database at `~/.local/share/codefence/history.db`.
+   - Directory mode `0o700`, file mode `0o600`.
+   - Contains: scan timestamp, tool version, file count, finding
+     counts, rule IDs, file paths, line numbers, and message text.
+   - Contains **no source code** and **no secret values**. Only
+     metadata about scans.
+   - Disable with `--history` omitted (default) or with
+     `CODEFENCE_NO_HISTORY=1` in the environment.
+   - Delete the directory at any time to clear history.
+
+5. **Distribution tampering**
+   - Published package includes `CHECKSUMS.txt` with SHA-256 of every
+     shipped file. Users are advised to verify after download.
+
+## Platform limitations and safe fallbacks
+
+- **Signal-based regex timeout.** On Unix-like systems and the main
+  thread, CodeFence enforces a hard 0.5-second timeout on every
+  regex call. On platforms where signals are unavailable (Windows)
+  or when running in a non-main thread (e.g. an async worker), the
+  timeout cannot be enforced by the OS. In that case CodeFence:
+  1. Prints a one-time warning to stderr.
+  2. Falls back to a strict input cap of 1024 characters per regex
+     call.
+  3. Continues the scan; results may be less complete on very long
+     lines.
+  This behavior is intentional: silent failure of ReDoS protection
+  would be worse than a visible warning.
+
+- **Baseline fingerprint includes line and column.** A finding's
+  fingerprint is derived from rule ID, file path, line, column, and
+  normalized snippet text. This means:
+  - Two findings of the same rule on different lines are distinct
+    (correct behavior — no collision).
+  - If code shifts up or down within a file (e.g. after adding
+    lines above), existing findings may appear as "new" in
+    `--diff`. This is intentional: catching new instances of a
+    dangerous pattern matters more than avoiding re-reporting a
+    shifted one.
+  - To suppress a known false positive permanently, use
+    `# noqa: RXXX` on that line, which is not affected by line
+    shifts.
+
+- **Policy self-modification.** A pull request can modify
+  `.codefence/policy.json` in the same commit as the code it
+  affects. This allows a PR to weaken its own gate. CodeFence
+  prints a warning when the policy or config file is staged in the
+  current commit, but does not block. Review policy changes
+  carefully in code review.
+
+## Supply chain
+
+This tool has **zero third-party dependencies**. It only uses modules
+from the Python standard library (`argparse`, `ast`, `hashlib`, `html`,
+`json`, `os`, `re`, `signal`, `stat`, `sys`, `tempfile`, `time`,
+`dataclasses`, `datetime`, `enum`, `pathlib`, `typing`,
+`unicodedata`).
+
+There is no `pip install`, no `npm install`, no vendored code, no
+build step, and no runtime download.
+
+## Reporting a security issue
+
+This is a one-time, as-is product. There is no official support
+channel. If you find a vulnerability you believe is worth sharing,
+please publish it responsibly — the maintainers do not operate a
+private disclosure inbox.
+
+## Honest limitations
+
+- A clean scan **does not mean the code is secure**.
+- The tool detects a curated set of dangerous **patterns**, not
+  vulnerabilities in general.
+- The tool does not perform dataflow or taint analysis.
+- False positives and false negatives are inherent to the approach.
+- Do not rely on this tool as the only security check before shipping
+  code. It is designed to be a fast, offline, first-pass filter.
+
+## License of the security model
+
+This document is provided for transparency. It is not a warranty.
+See `TERMS_OF_USE.md` for the legal terms.