feat(tools): get_temporal_priority — 0-100 CVE urgency scorer + CLI subcommand

[manus-use]

Summary

Implements the get_temporal_priority tool and manus-agent temporal-priority CLI subcommand — a 0–100 CVE urgency scorer that combines six signals to answer: "given everything I know today, how urgent is this CVE?"

This was documented in the README with full usage examples but had zero implementation (no tool file, no CLI dispatch, no tests).

Signals & Scoring

Signal	Weight	Logic
CVSS base score	25 pts	Linear: 0→0, 10→25
Current EPSS	25 pts	Non-linear (√): amplifies mid-range probabilities
EPSS spike recency	15 pts	Exponential decay (half-life 14d) from spike date
CISA KEV membership	20 pts	Binary: in catalog = full points
Patch availability	10 pts	Unpatched = +10 bonus; patched = −10 (reduces urgency)
CVE age	5 pts	Exponential decay (half-life 180d) from publish date

Final score normalized to 0–100 with labels: CRITICAL (≥85), HIGH (≥70), MEDIUM (≥50), LOW (≥30), INFORMATIONAL (<30).

Usage

manus-agent temporal-priority CVE-2024-3094
manus-agent temporal-priority CVE-2024-3094 --output json | jq .score

Design Decisions

• Zero new dependencies — uses only requests (already in deps)
• Graceful degradation — if any data source is unavailable, remaining signals still produce a partial score
• Retry/back-off on all HTTP calls (configurable via TEMPORAL_PRIORITY_MAX_RETRIES, TEMPORAL_PRIORITY_RETRY_DELAY)
• NVD_API_KEY support for higher rate limits
• Strands TOOL_SPEC interface — follows the module-based pattern used by other tools
• Configurable weights via environment variables (TEMPORAL_PRIORITY_W_CVSS, etc.)

Test Coverage

90 new tests (100% mocked, no real HTTP):

• TOOL_SPEC contract (4)
• Input validation (7)
• HTTP retry/back-off (5)
• NVD data fetching (5)
• CVSS extraction (5)
• Patch reference detection (5)
• EPSS data fetching (4)
• Spike analysis (6)
• KEV status checking (4)
• Score computation (9)
• Age calculation (6)
• Urgency labels (5)
• Text rendering (4)
• Integration pipeline (3)
• Handler (3)
• CLI subcommand (5)
• Safe float helper (5)
• Edge cases (6)

Full suite: 1248 passed, 0 failures (baseline 1158 + 90 new)

Duplicate Check

Checked all 50 open PRs (#136–#185) and 30 most recent merged PRs. No existing open or merged PR implements temporal-priority. Closest PRs:

• feat: add cve-enrich tool and CLI subcommand for multi-source CVE enrichment #170 (open, cve-enrich) — single-CVE enrichment dump, not a scoring algorithm
• feat(tools): EPSS trend tracker with spike detection + epss-trend CLI subcommand #45 (merged, epss-trend) — EPSS time-series only, no composite scoring
• feat(tools): exploit complexity scorer — score_exploit_complexity tool + manus-use exploit-complexity subcommand #57 (merged, exploit-complexity) — complexity scorer (1–5), different dimension entirely

Files Changed

• src/manus_agent/tools/get_temporal_priority.py (new) — tool implementation
• src/manus_agent/cli.py — added temporal-priority to _SUBCOMMANDS + parser/runner
• tests/test_temporal_priority.py (new) — 90 tests