The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes high-privilege shell commands to perform deep system analysis.
It instructs the use of sudo security add-trusted-cert to modify system-wide certificate trust on macOS, allowing for TLS decryption and interception via mitmproxy.
It uses sudo dtruss for system-level process tracing.
It uses sudo tee /proc/sys/kernel/randomize_va_space to modify kernel-level security settings by disabling ASLR.
It makes extensive use of powerful debugging and reverse-engineering tools like gdb, lldb, dlv, pdb, strace, and ltrace which can inspect and modify process memory and state.
[DATA_EXFILTRATION]: The skill's operational requirements involve accessing sensitive data with the ability to communicate over the network.
The methodology explicitly requires the agent to search for and read configuration files (.env, pyproject.toml, etc.), environment variables, and application/system logs.
It includes instructions for setting up network proxies (mitmproxy) and using network clients (curl, wget, httpie) to interact with endpoints, creating a potential path for data exfiltration.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core function of analyzing untrusted external data.
Ingestion points: The agent is instructed to read and analyze potentially attacker-controlled content from logs, binary strings, and web pages via browser automation.
Capability inventory: The agent possesses full shell access and the ability to execute high-privilege commands, meaning an injection could trigger severe system impact.
Sanitization: While the instructions call for verbatim recording of findings, there are no instructions to sanitize or escape data before the agent's internal reasoning engine processes it.
Boundary markers: No explicit boundary markers or warnings are used to prevent the agent from following instructions embedded in the analyzed data.
[EXTERNAL_DOWNLOADS]: The skill directs the agent to download and install a wide variety of third-party tools and utilities.
It provides installation instructions for complex software suites like Ghidra and pwndbg using package managers like brew, pip, npm, and cargo.
In references/runtimes/bundled-js-binary.md, it mentions using external scripts like pyinstxtractor.py and community-supported extraction tools.
It includes example patterns for fetching and inspecting remote installation scripts using curl.

debugging