ai-threat-testing
Offensive AI security testing and exploitation framework. Systematically tests LLM…
Generate ONE strong password and apply it to each referenced file (PDF, Word, Excel, PowerPoint, or any type). Use when asked to password-protect / encrypt / lock one or more deliverable files with a single password before sharing.
$ npx -y skills add transilienceai/communitytools --skill protect-with-password --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/protect-with-passwordContext preview
The summary Claude sees to decide when to auto-load this skill.
Generate ONE strong password and apply it to each referenced file (PDF, Word, Excel, PowerPoint, or any type). Use when asked to password-protect / encrypt / lock one or more deliverable files with a single password before sharing.
name: protect-with-password description: Generate ONE strong password and apply it to each referenced file (PDF, Word, Excel, PowerPoint, or any type). Use when asked to password-protect / encrypt / lock one or more deliverable files with a single password before sharing.
Mint a single strong password and password-protect every referenced file with it, "accordingly" per file type. Deterministic, non-destructive (originals are never touched), AES-256 or nothing.
python3 tools/protect_with_password.py FILE [FILE ...] \
[--password PW] [--length 24] [--out-dir DIR] \
[--password-file PATH | --no-password-file]| File | Protection | Result | |------|-----------|--------| | `.pdf` | Native AES-256 (qpdf → pypdf fallback) | `<name>-protected.pdf` — recipient is prompted for the password on open | | `.docx/.xlsx/.pptx`, `.doc/.xls/.ppt`, images, anything else | AES-256 7-Zip archive (`-mhe=on`, filenames encrypted) | `<name>.7z` — recipient extracts with the password (7-Zip / Keka / WinRAR) |
Native in-file Office (MS-OFFCRYPTO "agile") encryption needs LibreOffice or a dedicated encryptor that is not bundled here, so Office files are wrapped in a strong AES-256 archive instead. If no AES-256 tool is available for a file, it is reported as a failure — never downgraded to weak ZipCrypto/RC4 (no false assurance).
Relay the password to the user and remind them to **share it OUT-OF-BAND** — never in the same channel/email as the protected files — and to delete `PROTECT-PASSWORD.txt` once handed over. Exit code is 0 only if every file was protected; 1 if any failed (check `results[].reason`).
Related: [`tools/protect_deliverable.py`](../../tools/protect_deliverable.py) does the engagement-wide variant (one protected PDF + an AES-256 deliverable zip); this skill protects an arbitrary set of individual files.
Open-source Claude Code skills and agents for AI-powered penetration testing, bug bounty hunting, AI threat testing, and security reconnaissance — from the team at Transilience.ai
Repo: transilienceai/communitytools
Offensive AI security testing and exploitation framework. Systematically tests LLM…
API security testing - GraphQL, REST API, WebSocket, and Web-LLM attack techniques.
Stitches confirmed single-asset findings into multi-hop attack paths across the organization.…
Acquire an authenticated session THROUGH MFA/OTP on an in-scope target and emit a reusable…
Authentication security testing - auth bypass, JWT attacks, OAuth flaws, password attacks,…
Smart contract security testing and blockchain CTF exploitation. Covers Solidity…