More screenshots
Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+22 moreClaude CodeCursorCodex CLIMuseDots
OpenClaw
Manus
Grok
Grok Bot+18 agentsTimeForge: Deterministic Event Timeline & Causal Audit
Builds one UTC timeline from mixed logs, CSVs, Slack exports and git, then answers what-led-to-X with line-cited evidence and honesty flags.
$19
TimeForge: Deterministic Event Timeline & Causal Audit
Example session with this skill installed
Our checkout service paged at 14:04 and recovered by 14:15. I have the app log, gateway JSONL, deploy CSV, PagerDuty alert JSON, Slack export and the git log from that window in ./incident/. Build the timeline and show me the evidence chain around the page - tell me what actually connects to the 5xx spike and what is just nearby noise.
- Read your context and instructions
- Compiled the timeforge
$ node timeforge.mjs app.log gateway.jsonl deploys.csv alerts.json slack.json gitlog.txt
timeforge: 27 events (38 raw, 0 exact-dupes merged, 1 bursts collapsed)
span: 2026-10-17T08:30:00.000Z .. 2026-10-17T14:15:02.899902Z (5.751h)
kinds: change=4 note=8 error=4 burst=1 diagnosis=2 detection=4 mitigation=2 resolution=2
naive timestamps (assumedTZ tagged): 1
order-ambiguous second buckets: 4
warnings (8):
NON_MONOTONIC deploys.csv:L4 is earlier than deploys.csv:L3 in write order
CLOCK_SKEW gateway.jsonl: event 1500ms ahead of collector receipt
ASSUMED_TZ 1 event(s) had no explicit offset - tagged assumedTZ
ORDER_AMBIGUOUS 4 second-bucket(s) mix second and sub-second stamps
BURST app.log:L10..L21: 12 structurally-identical events in 0.3s
timeline written: timeline.json
$ node timeforge-query.mjs timeline.json chain E015 --window 300
chain for E015 @ 2026-10-17T14:02:05.700Z (gateway upstream timeout POST /v1/cart)
E008 [HARD_LINK shared trace_id=9f3c1a7b22] app ERROR upstream timeout GET /v1/orders cite: app.log:L3
E009 [HARD_LINK shared trace_id=9f3c1a7b22] gateway timeout GET /v1/orders cite: gateway.jsonl:L1
E014 [TEMPORAL_PROXIMITY +1.5s, no shared id] [BURST x12] Connection refused 10.0.1.4:5432 cite: app.log:L10
17 candidates fell in the window; 2 share the request's trace_id, 14 are temporal adjacency only, and the 12-event retry burst is one collapsed line. Deploy v2.14.0 sits 2m 5s before the first timeout with no shared identifier - flagged as proximity, not cause. The report.html output carries the same flags for the incident review doc.
Connects securely to your tools. The creator never sees your data.
About this skill
Most temporal-reasoning skills ship a prompt telling the agent how to think about time. The agent then has to sort 5,000 log lines inside its context window, guess at naive timestamps, and invent causality. TimeForge ships the engine instead.
What it actually runs
node timeforge.mjs app.log gateway.jsonl deploys.csv alerts.json slack.json gitlog.txt produces a single normalized timeline.json and a self-contained HTML report:
- Parses 9 timestamp shapes including ISO with ms/us precision, syslog dates, git log stamps, Slack decimal epochs, and Unix ms
- Merges exact duplicates and collapses structurally identical bursts (500 identical retries become one line with count and rate)
- Flags what it cannot prove: TZ-naive stamps, mixed second vs sub-second ordering, non-monotonic write order, clock skew
- Cites every event to its source file, line, or JSON path
node timeforge-query.mjs timeline.json chain E019 reconstructs the evidence chain around any event. Each link is labeled HARD_LINK (shared trace_id, request_id, span, commit or PR number) or TEMPORAL_PROXIMITY (adjacent in time, nothing else). A baseline-rate check separates NEW_EVENT and ANOMALOUS_SPIKE from BACKGROUND_NOISE so pre-existing error sources are not blamed for the outage.
What it refuses to do
- It never asserts causation. HARD_LINK means shared identifier, TEMPORAL_PROXIMITY means nothing else.
- It does not silently normalize ambiguous input. Naive timestamps are tagged assumedTZ, not rewritten.
- It emits no confidence scores, root-cause verdicts, or severity ratings. Those are judgement calls; this tool reports evidence.
Package contents
- 3 zero-dependency Node CLIs (timeline build, query, report)
- 20-test suite (node:test,
npm test) - Methodology and input-format reference docs
- Runnable example incident with real fixture inputs and captured output
Node 18+. Works with any SKILL.md-compatible agent.
How to use
Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scanned
Verified clean 2 days ago
- Passed all security checks, Safe to install
Requires
Git