The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external, potentially untrusted data sources including error logs, stack traces, HAR files, and captured network traces. * Ingestion points: The agent processes stack traces, log dumps, and network captures from the environment or user (SKILL.md Phase 1). * Boundary markers: No explicit delimiters or instructions to ignore embedded commands are specified for these inputs in the prompt construction. * Capability inventory: The agent is instructed to execute shell commands, git operations, network requests via curl, and run browser automation scripts (SKILL.md Phase 1). * Sanitization: The skill provides instructions to redact sensitive data (credentials, PII) but lacks technical sanitization for malicious instructions embedded in the analyzed artifacts.
[DYNAMIC_EXECUTION]: The debugging process encourages the creation and execution of temporary scripts and test harnesses to reproduce issues. * Evidence: Phase 1 recommends building 'Curl / HTTP script against a running dev server', 'CLI invocation with fixture input', and 'Throwaway harness (minimal subset of the system)' to create a feedback loop. * Evidence: Phase 3 involves 'Hypothesis and Testing' which may involve the generation of diagnostic code or probes.

oracle-debug