context-bundler — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited context-bundler (Agent Skill) and scored it 87/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 3 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 3 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
This skill requires Python 3.8+ and standard library only. No external packages needed.
To install this skill's dependencies:
pip-compile ./requirements.in
pip install -r ./requirements.txtThis skill centralizes the knowledge and workflows for creating "Context Bundles." These bundles compile large amounts of code and design context into either a single, portable Markdown file for sharing with other AI agents, or a compressed .zip file for native format sharing and human review.
Because context limits are strict and re-bundling is inefficient, this is a Level 2.0 Interactive Skill. You must follow the phased workflow below to confirm the target files and output format before generating the payload.
Discover, Confirm, and Package. You do not just "list files" or immediately run the bundling scripts. You ensure the bundle is targeted, complete, and annotated, getting user sign-off before execution.
When asked to bundle files, you MUST follow these phases in order. Do not skip to execution.
Evaluate the user's initial request. If it is vague (e.g., "Bundle the auth logic" or "Bundle these files"), ask targeted questions to shape the payload:
ls or codebase search to suggest 3-5 high-value files if they don't know)..md) to paste into an LLM, or a portable Archive (.zip)?Wait for the user's response before proceeding.
Draft the JSON manifest schema conceptually, but DO NOT execute the Python scripts or write to disk yet. Present the proposed plan to the user for approval:
Context Bundle Plan:
- Title: [Proposed Title]
- Format: [.md or .zip]
- Proposed Files/Directories:
1. src/main.py (Core logic)
2. docs/architecture.md (Design reference)
- Exclusions: (e.g., exclude .png, node_modules, or large JSON artifacts)
Does this look right? (yes / adjust / exclude certain extensions)Wait for the user to confirm.
Once confirmed, formulate the actual file-manifest.json on disk. IMPORTANT: Use directory paths (ending in /) to recursively include entire folders rather than listing 50 files individually.
{
"title": "Bundle Title",
"description": "Short explanation of the bundle's goal.",
"excludes": [
"**/large_artifact.json",
"**/*.png"
],
"files": [
{
"path": "docs/architecture.md",
"note": "Primary design document."
},
{
"path": "src/module/",
"note": "Implementation logic (recursive)"
}
]
}Invoke the appropriate script based on the format negotiated in Phase 1. (Adjust the script path below depending on if you are running this from the plugin root or via an npx installed `.agents/` path).
python ./scripts/bundle.py --manifest path/to/file-manifest.json --bundle path/to/output.md python ./scripts/bundle_zip.py --manifest path/to/file-manifest.json --bundle path/to/output.zipInform the user the payload is ready.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.