Add terraform-policy-controls plugin with terraform-runtask skill

[srlynch1]

Summary

• Adds a new terraform-policy-controls plugin under terraform/policy-controls/ with a terraform-runtask skill that fetches HCP Terraform/TFE run task stages, results, and outcomes via the REST API — filling a gap not covered by the Terraform MCP server.
• Registers the plugin in the top-level .claude-plugin/marketplace.json so it's discoverable alongside the other HashiCorp plugins.

The skill ships a shell script (scripts/get-run-task-results.sh) that authenticates with TFE_TOKEN, sideloads task results via include=task_results, then fetches each outcome's HTML body for rich reporting. The SKILL.md guides Claude to present results in a four-tier structure (summary → stages → task results → outcomes) and synthesize actionable findings from the HTML bodies.

Test plan

• jq . parses terraform/policy-controls/.claude-plugin/plugin.json and .claude-plugin/marketplace.json without error
• ./terraform/policy-controls/skills/terraform-runtask/scripts/get-run-task-results.sh run-<id> returns valid JSON with task_stages, summary, and per-outcome body_html content against a real HCP Terraform run
• Script exits cleanly on: missing TFE_TOKEN, invalid run ID format, 401/404 responses
• URL input form (.../runs/run-abc123) correctly extracts run ID and auto-detects TFE_HOSTNAME
• Plugin installs from the marketplace and the terraform-runtask skill activates on relevant prompts ("check the run tasks", "get task results for run-xxx")

[github-actions]

Tessl Skill Review Results

Skill	Status	Review Score	Change
terraform/policy-controls/skills/terraform-runtask	✅ PASSED	92%

---

Detailed Review

terraform/policy-controls/skills/terraform-runtask — 92% (PASSED)

  Description: 100%
    \(.key): \(.value.score)/3 - \(.value.reasoning)
    \(.key): \(.value.score)/3 - \(.value.reasoning)
    \(.key): \(.value.score)/3 - \(.value.reasoning)
    \(.key): \(.value.score)/3 - \(.value.reasoning)

    Assessment: This is a strong skill description that clearly defines its purpose, provides explicit trigger guidance, and uses domain-specific terminology that makes it highly distinguishable. It follows the recommended pattern of stating what it does followed by when to use it, with concrete example phrases that users would naturally say.

  Content: 83%
    \(.key): \(.value.score)/3 - \(.value.reasoning)
    \(.key): \(.value.score)/3 - \(.value.reasoning)
    \(.key): \(.value.score)/3 - \(.value.reasoning)
    \(.key): \(.value.score)/3 - \(.value.reasoning)

    Assessment: This is a well-crafted skill with excellent actionability and workflow clarity — it provides concrete commands, exact output formats, and thorough edge case handling. The main weaknesses are moderate verbosity (explanations of why certain design choices matter, which Claude doesn't need) and a monolithic structure that could benefit from splitting the field reference and presentation templates into separate files.

Suggestions:

• Trim explanatory rationale paragraphs (e.g., 'This distinction matters because...', 'This synthesis is what makes the skill output more valuable...') — Claude can infer the reasoning from the instructions themselves.
• Move the 'Reading the JSON output' field reference section to a separate REFERENCE.md file and link to it, reducing the main skill's token footprint.

---

Checks: frontmatter validity, required fields, body structure, examples, line count.
Review score is informational — not used for pass/fail gating.