The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes a surface for indirect prompt injection by ingesting and processing untrusted data from external Aircall webhook events.
Ingestion points: Webhook handlers in examples/nextjs/app/webhooks/aircall/route.ts, examples/express/src/index.js, and examples/fastapi/main.py ingest JSON payloads from external POST requests.
Boundary markers: The skill implements a shared secret token verification step using timing-safe comparisons (crypto.timingSafeEqual in Node.js and secrets.compare_digest in Python) to authenticate the sender.
Capability inventory: Example handlers primarily log the ingested event data to the console using console.log or logging.info. The skill does not demonstrate calling dangerous tools or executing commands directly on the untrusted input.
Sanitization: There is no explicit sanitization or filtering of the values within the data object before they are logged to system outputs.
[EXTERNAL_DOWNLOADS]: The skill instructions and configuration files reference external tools and dependencies.
Vendor Tooling: The README recommends using npx hookdeck-cli for local development to tunnel HTTPS traffic. This is documented as a legitimate tool provided by the skill's author (hookdeck).
[REMOTE_CODE_EXECUTION]: The manifest files reference unverifiable dependency versions.
Unverifiable Versions: The package.json files in the Express and Next.js examples, as well as the requirements.txt in the FastAPI example, specify version numbers for major packages that are significantly higher than currently available stable releases (e.g., Next.js 16.3.1, TypeScript 7.0.2, Vitest 4.1.10, Jest 30.4.2, FastAPI 0.141.1, and Pytest 9.1.1). Using non-existent versions in dependency manifests can lead to dependency confusion or the installation of malicious packages if an attacker registers these specific versions on public registries like npm or PyPI.

aircall-webhooks