The Agent Skills Directory

[SAFE]: No malicious patterns or high-severity security issues were detected. The skill's operations are consistent with its primary purpose of analyzing legacy codebases.
[PROMPT_INJECTION]: The skill contains operational instructions for framework phases (Archaeology, Resurrection, Rejuvenation). It does not attempt to bypass safety guidelines or override agent constraints. It presents a potential indirect prompt injection surface because it is designed to ingest and analyze untrusted legacy code.
Ingestion points: scripts/analyze-repo.sh and various discovery commands read content from repositories.
Boundary markers: None identified.
Capability inventory: Bash, Read, Write, WebFetch, Grep, Glob.
Sanitization: Content is typically processed via regex for report generation without specific sanitization filters.
[DATA_EXFILTRATION]: The discovery of environment variable names (e.g., process.env in scripts/analyze-repo.sh) and configuration paths is part of the repository archaeology process. No hardcoded credentials or unauthorized network exfiltration of sensitive data were found.
[COMMAND_EXECUTION]: Local bash scripts (analyze-repo.sh, scan-repos.sh) automate repository metadata analysis using standard system utilities like find, git, jq, and grep. These scripts do not attempt privilege escalation or persistence.
[EXTERNAL_DOWNLOADS]: The skill uses the GitHub CLI (gh) to list repository information from an organization, which is a standard administrative function for code auditing.

code-necromancer