The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill functions process untrusted data from external X/Twitter profiles and tweets, creating a potential ingestion surface for malicious instructions.
Ingestion points: Functions like viewJoinDate('username') and generateTweet('tweetUrl') fetch content from external profile and tweet pages (SKILL.md).
Boundary markers: The skill lacks explicit delimiters or instructions to ignore embedded commands within the ingested data.
Capability inventory: Capabilities include exporting full account summaries to JSON, navigating to sensitive settings pages, and uploading phone contacts.
Sanitization: No evidence of input validation or content filtering is present in the documentation for the external data.
[DATA_EXFILTRATION]: The skill defines functions that access highly sensitive Personal Identifiable Information (PII) and security data, such as viewLoginHistory (active sessions, device, location, IP), viewConnectedAccounts (Google/Apple links), and uploadContacts.
[METADATA_POISONING]: The author field in the metadata (nichxbt) does not match the expected skill author (nirholas), indicating potential misattribution or inconsistent documentation.
[NO_CODE]: The primary logic for the described tools resides in external files (src/accountMisc.js, src/qrCodeSharing.js, src/shareEmbed.js, src/uploadContacts.js) which were not provided for analysis, preventing a full audit of the execution logic.

account-tools