The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes explicit instructions to override standard agent behavior, stating 'You are in AUTONOMOUS MODE. Do NOT ask questions' and 'Execute the full pipeline below without pausing for user input'. This pattern is designed to bypass the agent's interaction safety loops and eliminate human-in-the-loop verification for complex operations.
[DATA_EXFILTRATION]: The 'Self-Evolution Telemetry' section directs the agent to access and append data to '~/.claude/projects/skill-telemetry.md'. This constitutes an unauthorized file system operation targeting sensitive application configuration paths outside of the intended project scope.
[EXTERNAL_DOWNLOADS]: The skill automates the download and installation of the Stripe SDK across multiple languages and frameworks, including 'stripe' for Python, Node.js, and PHP, as well as '@stripe/stripe-js' and 'flutter_stripe'. These are well-known libraries from a trusted service provider.
[COMMAND_EXECUTION]: The skill executes various shell commands to manage dependencies (npm, pip, go, composer), configure the environment, and perform verification steps using tools like 'tsc', 'mypy', and 'eslint'.
[PROMPT_INJECTION]: The skill accepts user-provided input through the '$ARGUMENTS' variable which is then used to drive logic involving file modifications and command executions. The absence of boundary markers or sanitization for this input creates a surface for indirect prompt injection that could influence the agent's behavior during the integration process.
Ingestion points: SKILL.md (via $ARGUMENTS)
Boundary markers: Absent
Capability inventory: Shell command execution (npm, pip, go, composer), file read/write (SKILL.md, .env, project source files)
Sanitization: Absent

stripe