The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where the agent is instructed to ingest and process CI/CD failure logs to diagnose and fix errors. If an attacker can influence the build output (e.g., through a malicious pull request), they could potentially embed instructions in the logs to redirect the agent's behavior.
Ingestion points: Section 'Feeding CI Failures Back to Agents' in SKILL.md.
Boundary markers: Absent; the example prompt lacks delimiters to distinguish the pasted error logs from instructions.
Capability inventory: The skill utilizes shell command execution via npm, npx, and Vercel tools, and references the 'incremental-implementation' skill for code modifications.
Sanitization: No sanitization or validation of the ingested log content is mentioned.
[METADATA_POISONING]: Several inconsistencies are present in the skill's metadata.
Author Discrepancy: The YAML frontmatter lists 'addyosmani' as the author, which contradicts the provided author context of 'seaworld008'.
Future Dating: The 'created_at' and 'updated_at' fields use dates in 2026, which is inconsistent with current system time.
[COMMAND_EXECUTION]: The skill relies on executing various development and deployment tools via the shell.
Evidence: Usage of npm ci, npm run lint, npx tsc, npm test, npx playwright, and npx vercel is documented throughout the SKILL.md templates.
[SAFE_PRACTICE]: The skill follows security best practices for secret management, explicitly advising against hardcoding credentials and recommending the use of GitHub Secrets or dedicated vaults.

ci-cd-and-automation