ops:full

Prompt file imported from jifflee/eagle-eye (.claude/commands/ops:full.md). Copyright stays with the author.

Ops Full Audit

Orchestrates all 6 ops skills, aggregates findings into a unified report, and submits results as a GitHub issue to jifflee/claude-tastic. Large payloads are compressed via gzip+base64 to stay within GitHub's ~65K character issue body limit.

READ-ONLY OPERATION — collects data from read-only ops skills; the only write action is issue creation in the target repo.

Usage

/ops:full                          # Run all audits and submit issue (default)
/ops:full --no-submit              # Run all audits, print report only
/ops:full --dry-run                # Preview issue body without submitting
/ops:full --repo owner/repo        # Submit to different repo (default: jifflee/claude-tastic)

Arguments

Steps

1. Parse Arguments

Check for flags and set defaults:

• TARGET_REPO="jifflee/claude-tastic" (override with --repo)
• SUBMIT=true (set to false with --no-submit)
• DRY_RUN=false (set to true with --dry-run)

2. Run All Ops Skills via Skill Tool

Invoke each skill sequentially and capture output. Use the Skill tool for each:

1. /ops:metrics — token usage, model selection, performance
2. /ops:skills — skill token efficiency and scripting opportunities
3. /ops:agents — agent format compliance and quality
4. /ops:actions — action audit log and tier analysis
5. /ops:models — model configuration distribution
6. /ops:skill-deps — dependency visualization and tier inheritance

Store each output with a section label for aggregation.

3. Aggregate into Unified Report

Combine all outputs into a structured markdown report:

# Ops Full Audit Report

**Generated:** {ISO-8601 timestamp}
**Repository:** {current repo name via `gh repo view --json name -q .name`}
**Framework:** claude-tastic consumer ops health check

---

## Executive Summary

| Skill | Status | Key Finding |
|-------|--------|-------------|
| ops:metrics | {OK/WARN/CRIT} | {one-line summary} |
| ops:skills | {OK/WARN/CRIT} | {one-line summary} |
| ops:agents | {OK/WARN/CRIT} | {one-line summary} |
| ops:actions | {OK/WARN/CRIT} | {one-line summary} |
| ops:models | {OK/WARN/CRIT} | {one-line summary} |
| ops:skill-deps | {OK/WARN/CRIT} | {one-line summary} |

---

## ops:metrics

{full output}

---

## ops:skills

{full output}

---

## ops:agents

{full output}

---

## ops:actions

{full output}

---

## ops:models

{full output}

---

## ops:skill-deps

{full output}

---

## Recommendations

{cross-cutting recommendations from all findings}

4. Apply Compression if Needed

Check report length and apply the framework compression standard if needed:

REPORT_LEN=${#REPORT}
LIMIT=60000   # Safe threshold below GitHub's 65,535 char limit

if [ "$REPORT_LEN" -gt "$LIMIT" ]; then
    # Compress: gzip + base64
    COMPRESSED=$(echo "$REPORT" | gzip -9 | base64 -w 0)
    ISSUE_BODY="$(cat <<'BODY'
# Ops Full Audit Report (Compressed)

> **Payload size exceeded GitHub limit. Full report is gzip+base64 encoded below.**
> **Compression standard:** See [docs/COMPRESSION_STANDARD.md](../docs/COMPRESSION_STANDARD.md)

## Decompression Instructions

Save the payload below to a file, then run:

\`\`\`bash
# Decompress on Linux/macOS
echo "<PAYLOAD>" | base64 -d | gunzip > ops-full-report.md
cat ops-full-report.md

# Or pipe directly:
echo "<PAYLOAD>" | base64 -d | gunzip | less
\`\`\`

## Executive Summary

{executive_summary_section}

## Compressed Payload

\`\`\`
<PAYLOAD>
\`\`\`

BODY
)"
    ISSUE_BODY="${ISSUE_BODY//<PAYLOAD>/$COMPRESSED}"
else
    ISSUE_BODY="$REPORT"
fi

Refer to docs/COMPRESSION_STANDARD.md for the canonical compression algorithm and decompression instructions.

5. Submit Issue to Target Repo

Unless --no-submit or --dry-run:

REPO_NAME=$(gh repo view --json name,owner -q '"\(.owner.login)/\(.name)"' 2>/dev/null || echo "unknown/repo")
DATE=$(date -u +%Y-%m-%d)

gh issue create \
  --repo "$TARGET_REPO" \
  --title "ops:full audit — $REPO_NAME ($DATE)" \
  --label "ops-report,automated" \
  --body "$ISSUE_BODY"

For --dry-run, print $ISSUE_BODY and the character count without submitting.

Output Format

Success

## Ops Full Audit Complete

**Skills run:** 6/6
**Report size:** {N} characters ({compressed: yes/no})
**Issue submitted:** {URL or "not submitted (--no-submit)"}

{executive summary table}

Run with --no-submit to print the full report locally.

Dry Run

## Ops Full Audit — Dry Run

**Report size:** {N} characters
**Compression:** {applied/not needed} (limit: 60,000 chars)
**Would submit to:** {TARGET_REPO}

--- ISSUE BODY PREVIEW ---

{issue body}

Token Optimization

• Each sub-skill already uses data scripts (pre-computed JSON)
• Aggregation is additive — no re-analysis needed
• Compression offloaded to bash (gzip, base64) — zero Claude tokens
• Estimated total: ~3,000–5,000 tokens across all 6 sub-skills

Notes

• Compression standard: docs/COMPRESSION_STANDARD.md — defines the canonical gzip+base64 method for all framework skills
• Target repo: Default jifflee/claude-tastic; requires gh auth with repo write access to that org
• Labels: Creates ops-report and automated labels on the target repo if they do not exist
• Consumer repo usage: Run from any consumer repo — the report includes the repo name so findings are attributed correctly
• Sub-skill order: Sequential to avoid rate-limit collisions; each sub-skill is READ-ONLY
• NEVER invoke /sprint-work or write operations from this skill