The Agent Skills Directory

[DATA_EXFILTRATION]: The documentation link https://jeffallan.github.io/claude-skills/skills/quality/debugging-wizard/ in SKILL.md is explicitly flagged as malicious and blacklisted by automated scanners. The inclusion of this URL in the skill's instructions poses a high risk if the agent is directed to interact with the domain or exfiltrate data to it.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data.
Ingestion points: The skill parses error messages, stack traces, and logs as defined in SKILL.md and references/systematic-debugging.md.
Boundary markers: There are no instructions for the agent to use delimiters or ignore embedded instructions within these data sources.
Capability inventory: The agent is instructed to execute shell commands for debuggers (pdb, node --inspect, dlv) and automated tests (git bisect run) across multiple files.
Sanitization: There is no evidence of input filtering or escaping for external content.
[REMOTE_CODE_EXECUTION]: Automated reputation scanners identified SKILL.md as containing FileRepMalware. The skill also recommends installing unversioned packages (e.g., npm install x in references/quick-fixes.md) and uses third-party Go libraries (github.com/davecgh/go-spew/spew in references/debugging-tools.md) which increase the attack surface for supply chain compromises.
[COMMAND_EXECUTION]: The skill facilitates the execution of arbitrary shell commands, including attaching debuggers to running processes and running regression tests (git bisect run npm test) as seen in references/strategies.md. When these capabilities are combined with the ingestion of untrusted log data, it creates a risk for command injection and unauthorized execution.

debugging-wizard