AI Agent Workflow Collection

Fault diagnosis and response playbook for Seestar S50 sessions. Use when something goes wrong mid-session or a check fails — stacking count stalls, rejection rate spikes, plate-solve fails, autofocus fails, focus drifts, the mount loses tracking, the connection drops, or QA shows a sudden quality collapse. Maps each symptom to likely causes and a concrete response, and says when to act automatically vs when to surface a decision to the user. The run-session skill routes faults here.

The processing RECIPE for turning Seestar OSC subs into a finished image with the pure-Python AstroPipe pipeline — the correct calibration/stacking/stretch ORDER and the object-type-specific parameters. Use whenever choosing pystack + stretchmaster params, tuning a result, or diagnosing artifacts (rotation "fans", blown cores, noisy backgrounds, small off-center galaxies). Complements image-refinement (which picks the backend); this skill decides the PARAMETERS. Grounded in the research synthesis at docs/2026-07-06-fan-artifact-research.md.

Unattended full-night run-book for a Seestar S50: assess, run the ranked plan target-by-target, react to conditions/QA, and wind down + park at dawn — on its own, behind hard safety guardrails and a mandatory dry-run confirmation. Use whenever the user wants to hand over the whole night — e.g. "run the whole night", "image unattended", "run an autonomous session", "image all night on your own", "run my target list unattended", "let it run itself till dawn". Starts with a NO-MOTION simulatenight projection that the user must explicitly approve before the first motion command. Orchestrates the existing tools in a visible loop — it decides whether to keep going and what's next; it does not re-implement motion (that is run-session).

Turn a session's QA keep-list into a finished image: stack the good subs into a master and refine them. Use whenever the user wants to process/refine a session — e.g. "stack my subs", "process the session", "refine the images", "make a final image", "create the final photo", "give me a finished picture of M27". Refines ONLY the QA keep-list (never rejected subs), stacks with DeepSkyStacker (default, always-available → master + auto-stretched preview), the pure-Python pystack backend (no external app), or, if the user has PixInsight, via WBPP + a hand-off to their external pixinsight-mcp. A pure-Python post-processing pipeline (gradient removal, color calibration, deconvolution, saturation, opt-in labeled AI upscale) is available via preview params. Runs on the seestar-refine MCP service; every run is provenance-logged.

Pre-session planner for a Seestar S50 night: is it worth imaging, and what should you shoot? Use whenever the user wants to plan an observing night — e.g. "plan tonight", "what should I image tonight", "is it clear enough to image", "build a target list", "pick targets for tonight", "give me a shortlist for the Seestar". Produces a one-line go/no-go conditions verdict and a ranked, reasoned top-3 target shortlist (best window, recommended integration, and the "why"). Consults the planning tools for every number — never invents ephemeris or weather — and hands the chosen target to the run-session skill for execution.

Scoring policy for Seestar S50 sub-frame quality — how to interpret the metrics returned by qatier2 (FWHM, HFR, eccentricity, SNR, background, star count, wFWHM) and decide PASS / MARGINAL / REJECT, plus how to read a qasessionreport. Use whenever judging whether collected data is "good", deciding which subs to keep for stacking, explaining why a sub was rejected, or setting/adjusting quality thresholds. This skill owns the numbers; the run-session skill calls it for interpretation.

Run-book for executing a Seestar S50 imaging session via the seestar-mcp tools. Use whenever the user wants to start, monitor, or wind down an imaging session — e.g. "image the Veil Nebula tonight", "start a session on M31", "slew to NGC 7000 and start stacking", "wrap up the session and pull the subs". Covers pre-flight checks, target acquisition, autofocus, stacking, in-session monitoring cadence, and clean shutdown. Defers QA scoring decisions to the qa-policy skill and fault response to the anomaly-playbook skill.