Works with the AI Agents you already use
repo context pack
Safely packages a repository into one LLM-ready context file, skipping sensitive and irrelevant files, scanning included content for common credential patterns, and requiring a preview before writing output.
$7
repo context pack
Example session with this skill installed
Pack the StartAI repository into a context file for an AI agent performing an architecture and security review. First run a full default-scope dry run to scan supported files for credential-looking text. If the scan is clean, run a second dry run including Python, Markdown, JSON, YAML, TXT, HTML, and CSS files while excluding the templates and feed directories. Do not write the final context file until I approve the preview.
- Read your context and instructions
- Compiled the repo context
I will first run a full default-scope dry run without writing any file. This scans the supported repository files for common credential patterns before narrowing the scope.
The script does not execute project code, make network requests, or print file contents.
DRY RUN, nothing written (would write /tmp/startai_context.txt)
Source: git-tracked files | included: 76 files, 396461 characters
skipped (hidden): 35
skipped (extension): 17
The full-scope dry run completed without credential-looking findings.
For the architecture and security review, I will now prepare a narrower second preview containing Python, Markdown, JSON, YAML, TXT, HTML, and CSS files while excluding the templates and feed directories:
python3 /path/to/repo-context-pack/scripts/context_pack.py \
--root /work/startai \
--extensions .py,.md,.json,.yaml,.yml,.txt,.html,.css \
--ignore-dirs templates,feed \
--dry-run
DRY RUN, nothing written (would write /tmp/startai_context.txt)
Source: git-tracked files | included: 35 files, 237619 characters
skipped (hidden): 35
skipped (ignored dir): 50
skipped (extension): 8
The scoped preview also completed without credential-looking findings. No final context file has been generated yet.
The selected context contains approximately 238,000 characters, leaving room for an architecture or security-review request and the agent’s response. After approval, I can run the same command without --dry-run and write the result outside the repository.
The generated file will contain a file list followed by the selected file contents. It should be reviewed before sharing. The credential scan is pattern-based and is not a complete secret scanner; skipped, hidden, binary, oversized, and unsupported files are not included or scanned.
Connects securely to your tools. The creator never sees your data.
About this skill
Giving a large codebase to an LLM can hit context limits or accidentally expose credentials. This skill flattens selected repository files into a single context-ready text file and runs a pattern-based credential scan that aborts when it detects suspicious content.
What it does
- Codebase flattening combines selected repository files into one context-ready document with clear file headers.
- Credential scanning detects private-key blocks, JWT-looking strings, and common API token and secret-assignment patterns.
- Smart filtering skips hidden files, binaries, symlinks, secret-looking filenames, oversized files, unsupported extensions, and common generated directories such as
node_modules. - Atomic file writing prevents partial output files and aborts before writing when a credential-looking finding is detected, unless the user explicitly approves that exact finding or skips the flagged file.
- Scope control lets users narrow the pack by extension, directory name, filename pattern, or maximum file size.
How it works
- Run a dry run to preview the size and scan the selected files without writing the output file.
- If the scan finds a possible credential, review the masked finding and decide whether to fix it, approve that exact line, or leave the flagged file out.
- Refine the scope with extensions, ignored directories, exclusions, or file-size limits when needed.
- Run the command without
--dry-runto write the pack outside the repository.
Frameworks & tools
This skill uses a standalone Python script compatible with Python 3.8 or newer. It uses only the standard library, makes no network requests, and never executes project code. It uses git ls-files when the repository is under Git; otherwise it walks the filesystem.
Why use this instead of prompting manually?
Manual copy-pasting is slow and can truncate files or miss important context. This skill automates repository selection and packaging while adding a repeatable, pattern-based safety check for common credential formats.
Use cases
- Preparing repository context for an architecture or code review with an LLM.
- Sharing selected code without including common secret files such as
.env. - Auditing which files an AI agent would receive as repository context.
- Reducing context size by excluding irrelevant, generated, binary, or oversized files.
The scan is pattern-based and is not a complete secret scanner. It may miss encoded, split, generated, or unknown credential formats and may produce false positives. Files skipped because they are hidden, binary, oversized, unsupported, unreadable, or secret-looking are not included or scanned. In Git repositories, untracked files are excluded by default unless --all-files is used. A clean result does not prove that the output contains no secrets.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 day ago
Requires