v0.3.2: Leaner guidance, safer client, RSS Auto Sync

Changes

• Cut the skill and reference text by about 48%, removing repetition and incorporating lessons from recent tool work.
• Corrected guidance on credit resets, upload limits, document pagination and attachments, image output, and Message Insights. Added clearer limits for paid testing and evidence-based evaluation.
• Hardened the Python HTTP client: no automatic retries by default, strict named-workspace credentials, better error detection, and JSON/MCP SSE response handling. Added 16 offline regression tests.
• Documented daily RSS Auto Sync when enabled. New feed items append to the knowledge base, and URLs remain after they leave the feed window. The guide distinguishes this from the older website-source workaround.

Validation

All 16 offline client tests and skill validation passed. This update did not include live API compatibility tests or paid completion runs.

Updating

Run git pull in your existing skill clone, or install from the repository README. You can pin this version with the v0.3.2 tag.

v0.3.1

A measured trigger description, the HTTP fallback reframed, and two API facts.

• The SKILL.md description was chosen by a trigger optimizer against twenty realistic prompts, isolated from any installed copy of the skill. It passed 7 of 8 held-out prompts with zero false triggers on near-misses (the npm package called pickaxe, Minecraft, other chatbot platforms), against 4 of 8 for the v0.3.0 text. The prompts and the harness caveats are in evals/.
• The JSON-RPC fallback is the remedy for any MCP client-layer failure, not only large payloads. An InputValidationError or parse error is deterministic, so never resend it unchanged.
• user_update rejects uses[] with 400, so per-deployment enforcement counters are read-only over the API.
• The deployment record carries its cap as usageLimit, where -1337 means unlimited, verified live.

v0.3.0

Twenty-three field lessons merged, a new grading-and-fixtures reference split out of testing-and-verification, and a fix to the bundled client.

• New reference: references/grading-and-fixtures.md covers rubrics, test fixtures, run counts, ground-truth grading, measurement validation, and anti-hallucination design. references/testing-and-verification.md keeps staging, transports, write verification, and the platform signals that mislead diagnosis. Every long reference now opens with a section list.
• Three corrections to published text: the answerlength co-update rule does not apply to upload fields, the UI-only class is per field (descriptions save, submittext does not), and a workspace fallback can change reasoning effort as well as model.
• New platform facts: pickaxe_create defaults, chat history as a recent window, action_runs carrying no duration, image action files as public CDN URLs, the three Access denied surfaces, documentuploadtype, stale sandboxes with zero attachments, sibling-tool drift, the frame as a sentence, truncation fabricating content, a Model Reminder measuring as a null, admin group toggles destroying purchased credits, and the public completions endpoint exercised with stream, inputs, and imageUrls.
• scripts/pickaxe_client.py: run_completion now takes inputs as well as message, and unwraps the completion whether it arrives as a Python-repr string or a structured dict, which the platform now returns.

v0.2.1

README rewrite, aimed at builders who have never used an agent skill before. Plain-words explanation of what a skill is, five concrete failure stories instead of a topic list, per-assistant install commands, and a Pickaxe Skill wordmark. No reference content changed from v0.2.0.

v0.2.0

Second release, two days after launch.

Added since v0.1 (the initial publish)

• Two more clients: install paths and MCP connection recipes for Grok Build and Grok Bot, alongside Claude Code, Codex, and Cursor.
• 40+ new field-tested lessons merged from production closeouts, including the message-vs-inputs transport distinction with a canary-probe technique, null-budget upload starvation, the chatflag/type coupling, promotion-carries-config, the silent workspace fallback model, UI-only fields, and SSO identity mechanics.
• Corrections: the staging checkout now includes the prompt frame and token budgets, verification compares behavior-determining fields rather than copied ones, and the overview no longer claims everything in the Studio UI is scriptable.

Evaluated

This release was graded with Warp's open-source skill-doctor against 11 real agent sessions (B+, 0.89 overall), and every drafted improvement from that evaluation is included.

Everything remains dated observations from the live platform. When the platform contradicts this skill, trust the platform and open an issue.