pixiv

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs the agent to ask the user for their Pixiv refresh token and to include that token verbatim in a CLI command (node ... login <REFRESH_TOKEN>) and to copy tokens from browser storage, which requires the LLM to handle and output secrets directly.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow for this skill (e.g., search, ranking, user, feed, following, download) calls Pixiv APIs and then prints/returns the resulting JSON (including titles/user text) into the agent context, which is outsider-authored free text originating from Pixiv users and public Pixiv content.

HIGH W008: Secret detected in skill content (API keys, tokens, passwords).

• Secret detected (high risk: 1.00). I scanned the skill files for high-entropy literal values that could be used to access services. I ignored documentation placeholders (e.g., "YOUR_PIXIV_REFRESH_TOKEN") and example/simple passwords (none present). I found a high-entropy hex constant labeled as a reverse-engineered "SALT" in two scripts. This salt is not a trivial placeholder and is used to compute X-Client-Hash headers (i.e., part of the client authentication/emulation logic), so it is a literal, high-entropy value that could be used to impersonate the app when combined with other requests — therefore it qualifies as a secret-like value per the definition.

Files and exact value:

• scripts/pixiv-app-publish.js: const SALT = "28c1fdd170a5204386cb1313c7077b34f83e4aaf4aa829ce78c231e05b0bae2c";
• scripts/pixiv-cli.js: const SALT = "28c1fdd170a5204386cb1313c7077b34f83e4aaf4aa829ce78c231e05b0bae2c";

Ignored items:

• config.json example uses "YOUR_PIXIV_REFRESH_TOKEN" — documentation placeholder, ignored.
• No refresh tokens, API keys, private keys, or other high-entropy tokens were present.