The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill retrieves scan definitions from https://api.appstorereject.com/api/scan/checks and instructs the agent to execute instructions found in the executionRule field against the project files. This allows external logic to drive local filesystem analysis and operations.
[DATA_EXFILTRATION]: Project metadata including the bundleId, framework, and platform are sent to the vendor. Additionally, scan findings containing up to 200 characters of code context are transmitted to https://api.appstorereject.com/api/scans/complete during the reporting phase, which contradicts the skill's privacy claims.
[COMMAND_EXECUTION]: The skill generates and executes shell commands dynamically. Step 8 requires the agent to execute a fetchCommand string constructed at runtime by a script. Step 0 also directs the agent to manually interpolate its API key into curl command strings, potentially exposing secrets in command logs.
[EXTERNAL_DOWNLOADS]: The skill performs multiple network requests to fetch configurations, rule graphs, and resolution guides from api.appstorereject.com.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by ingesting untrusted content (resolution guides and finding templates) from a remote API and requiring the agent to present this data verbatim or use it as guidance for searches without sanitization.
Ingestion points: asr-checks.json (Step 5) and asr-guides.json (Step 10)
Boundary markers: Absent; instructions explicitly mandate verbatim display of remote content.
Capability inventory: curl, grep, find, and standard file read/write operations.
Sanitization: Absent for finding templates and resolution guides.
[CREDENTIALS_UNSAFE]: The skill reads an API key from a configuration file at ~/.appstorereject/config.json and instructs the agent to store it in memory for manual insertion into shell commands.

appstorereject-scan