The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local CLI tools including qodo, git, and gh. It follows defensive patterns by resolving absolute paths, avoiding unsafe script launchers on Windows (like .bat or .ps1), and performing version compatibility checks before use.
[EXTERNAL_DOWNLOADS]: The skill provides procedures for updating the qodo CLI and installing verified updates. It explicitly mandates the use of checksum-pinned installers and forbids the unsafe practice of piping remote scripts directly into a shell.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from pull requests and review findings, which are externally controlled inputs. This presents a potential surface for indirect prompt injection if malicious instructions are embedded in the code or comments being reviewed.
Ingestion points: PR URLs and structured review session findings (JSON) provided by the Qodo tools.
Boundary markers: The instructions direct the agent to evaluate findings against the code and PR intent, though specific technical delimiters for data processing are not detailed.
Capability inventory: File system writes (local code fixes), network operations via the qodo CLI, and shell execution of version control tools.
Sanitization: The skill uses structured JSON tools which reduces parsing vulnerabilities, but lacks explicit sanitization instructions for the natural language content within those findings.

qodo-review-resolver