The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, such as Discord/Slack exports, Discourse forum data, and git history, which creates a potential surface for indirect prompt injection attacks.
Ingestion points: SKILL.md identifies external data sources including venue analytics, forge API, and git history in the "Instrumentation reality check" section.
Boundary markers: The skill instructions in SKILL.md recommend filtering bots and employee accounts but do not specify the use of clear delimiters (e.g., XML tags or custom markers) to separate external data from system instructions.
Capability inventory: The skill provides the agent with the capability to execute the scripts/contributor-absence-factor.sh bash script and references other skills for broader DevRel metrics.
Sanitization: SKILL.md focuses on bot filtering and identity mapping to ensure metric accuracy, but lacks explicit guidance on sanitizing text content to prevent adversarial instruction following during analysis.
[COMMAND_EXECUTION]: The skill utilizes a local bash script to perform git repository analysis, which involves the execution of standard shell utilities.
Evidence: SKILL.md instructs the agent to utilize scripts/contributor-absence-factor.sh. The script file itself uses git log, grep, sort, uniq, and awk to compute the contributor absence factor based on local repository data.

developer-community-health