generated: '2026-08-13' method: searched source: >- https://github.com/pulsemcp/ai-artifacts — the provider's own AIR-shaped catalog of agent artifacts. skills/skills.json was saved verbatim to skills/pulsemcp-skills.json on 2026-08-13; each SKILL.md lives at the raw URL recorded on its entry below. description: >- PulseMCP publishes 18 Agent Skills of its own, indexed in an AIR skills.json and authored as SKILL.md files. They are NOT skills for calling the PulseMCP Registry API — no such skill exists, and there is no OpenAPI to ground one in. They are a single plugin, `agent-transcript-analysis`: a four-tier pipeline that ingests Claude Code session transcripts, decomposes them into a segment tree, analyses prompting / Skills / MCP performance per segment, and synthesizes an actionable report. This is PulseMCP's agentic-engineering consulting practice packaged as artifacts. NOTHING HERE WAS AUTHORED BY API EVANGELIST. Every entry is the provider's own title, description and path, transcribed from their index. status_caveat: >- The repository README states plainly: "Status: scaffold. The skills capture the intended flow but are not implemented yet." Recorded as published-but-scaffold — the artifacts exist and are real, the implementations behind them may not. catalog: repo: https://github.com/pulsemcp/ai-artifacts format: AIR (Agent Infrastructure Repository) root_config: https://github.com/pulsemcp/ai-artifacts/blob/main/air.json index: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/skills.json index_schema: https://raw.githubusercontent.com/pulsemcp/air/main/schemas/skills.schema.json saved_verbatim: skills/pulsemcp-skills.json plugin: agent-transcript-analysis plugin_description: >- Inspect, redact, and analyze Claude Code session transcripts to surface actionable advice on prompting, Skills, and MCP servers. companion_artifacts: hooks: - id: agent-transcript-capture url: https://github.com/pulsemcp/ai-artifacts/tree/main/hooks/agent-transcript-capture description: Capture-side hook that archives agent transcripts, with inline redaction. references: - id: open-transcripts url: https://github.com/pulsemcp/ai-artifacts/tree/main/references/open-transcripts description: >- A vendor-neutral data model for coding-agent transcripts — schemas for the Transcript wrapper, events and the Transcript Segment, a Claude Code JSONL mapping, and minimal examples. - id: philosophy-on-skills url: https://github.com/pulsemcp/ai-artifacts/blob/main/references/philosophy-on-skills.md - id: philosophy-on-mcp url: https://github.com/pulsemcp/ai-artifacts/blob/main/references/philosophy-on-mcp.md - id: philosophy-on-prompting url: https://github.com/pulsemcp/ai-artifacts/blob/main/references/philosophy-on-prompting.md skill_count: 18 skills: - id: find-all-claude-code-transcripts-on-local title: "Find All Claude Code Transcripts On Local" path: agent-transcript-analysis/1-acquire/find-all-claude-code-transcripts-on-local url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/1-acquire/find-all-claude-code-transcripts-on-local/SKILL.md tier: '1' description: "List all Claude Code session transcripts on this machine and spawn a local browser UI for picking one to analyze. Privacy-preserving: server-side redaction, localhost only, no uploads." - id: get-claude-code-transcript-from-local title: "Get Claude Code Transcript From Local" path: agent-transcript-analysis/1-acquire/get-claude-code-transcript-from-local url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/1-acquire/get-claude-code-transcript-from-local/SKILL.md tier: '1' description: "Given a Claude Code session id (or a JSONL path), produce a single OpenTranscripts transcript.json \u2014 the main session plus every subagent it spawned, linked and embedded in one self-contained document. Owns the deterministic CC\u2192OT mapping: no LLM, no heuristics, secret-redaction inline." references: - open-transcripts-transcript - open-transcripts-events - open-transcripts-claude-code-mapping - id: gather-agent-transcript-external-context title: "Gather External Context" path: agent-transcript-analysis/1-acquire/gather-agent-transcript-external-context url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/1-acquire/gather-agent-transcript-external-context/SKILL.md tier: '1' description: "For a given transcript, gather the external context a reviewer would want before judging the session: the ticket the work traces back to (e.g. from Jira), the pull request it landed in (e.g. from GitHub), and background on the user's role/team/project. Consolidates everything into one external-context.json that travels with the transcript through every later tier. Best-effort: missing sources are recorded, never fatal." references: - open-transcripts-transcript - id: decompose-agent-transcript-into-transcript-segments title: "Decompose Agent Transcript Into Transcript Segments" path: agent-transcript-analysis/2-decompose/decompose-agent-transcript-into-transcript-segments url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/2-decompose/decompose-agent-transcript-into-transcript-segments/SKILL.md tier: '2' description: "Walk an OpenTranscripts transcript.json once and produce the recursive Transcript Segment tree (segments.json) plus an annotated flamegraph. Sole producer of the Segment primitive consumed by every downstream analyzer." references: - transcript-segment - open-transcripts-transcript - open-transcripts-events - id: analyze-agent-transcript title: "Analyze Agent Transcript" path: agent-transcript-analysis/3-analyze/analyze-agent-transcript url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/3-analyze/analyze-agent-transcript/SKILL.md tier: '3' description: "Orchestrator and entry point of the analyze tier (tier 3). Picks up the Segment tree from decomposition (tier 2), drives the per-Segment analyzers across four buckets, and writes that transcript's findings..json set \u2014 then stops. Runs per transcript; produces no report and never invokes synthesize-agent-transcript-analysis-report. Tiers 1-3 repeat per transcript and the findings sets accumulate; turning them into the consolidated report is tier 4's synthesize-agent-transcript-analysis-report, run once over the whole batch." references: - philosophy-on-skills - philosophy-on-mcp - philosophy-on-prompting - transcript-segment - id: analyze-agent-transcript-failure-hypothesis title: "Analyze Failure Hypothesis" path: agent-transcript-analysis/3-analyze/analyze-outcomes/analyze-agent-transcript-failure-hypothesis url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/3-analyze/analyze-outcomes/analyze-agent-transcript-failure-hypothesis/SKILL.md tier: '3' description: "Per-Segment analyzer. Produces an improvement hypothesis for every Failure Outcome and every retro-Failure (Correction Prompt at the next Segment's head). Routes its recommendation seed to the right downstream bucket." references: - philosophy-on-skills - philosophy-on-mcp - philosophy-on-prompting - transcript-segment - id: analyze-agent-transcript-segment-efficiency title: "Analyze Segment Efficiency" path: agent-transcript-analysis/3-analyze/analyze-outcomes/analyze-agent-transcript-segment-efficiency url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/3-analyze/analyze-outcomes/analyze-agent-transcript-segment-efficiency/SKILL.md tier: '3' description: "Per-Segment efficiency analyzer. Compares wall-clock / tokens to a human counterfactual and flags wasteful branches and model-tier mismatches \u2014 including on Success Outcomes." references: - transcript-segment - id: analyze-agent-transcript-user-prompt title: "Analyze User Prompt" path: agent-transcript-analysis/3-analyze/analyze-prompts/analyze-agent-transcript-user-prompt url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/3-analyze/analyze-prompts/analyze-agent-transcript-user-prompt/SKILL.md tier: '3' description: "Per-Prompt analyzer. Classify the Prompt (question vs delegation), confirm the Segment's Goal, judge whether it was closed-loop, and feed the human-prompting recommendation bucket. Consumes Initial/Correction classification from Tier 2." references: - philosophy-on-prompting - transcript-segment - id: analyze-agent-transcript-prompt-ambition title: "Analyze Prompt Ambition" path: agent-transcript-analysis/3-analyze/analyze-prompts/analyze-agent-transcript-prompt-ambition url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/3-analyze/analyze-prompts/analyze-agent-transcript-prompt-ambition/SKILL.md tier: '3' description: "Per-Initial-Prompt analyzer. Flags Initial Prompts that under-scoped the work and Initial Prompts that should have been deterministic triggers reacting to external events." references: - philosophy-on-prompting - transcript-segment - id: pull-together-agent-transcript-goal-context title: "Pull Together Goal Context" path: agent-transcript-analysis/3-analyze/analyze-prompts/pull-together-agent-transcript-goal-context url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/3-analyze/analyze-prompts/pull-together-agent-transcript-goal-context/SKILL.md tier: '3' description: "Reach into git repos and external systems when a Prompt's Goal is not self-evident from the Prompt alone. Ground the analyzer's understanding before drawing conclusions." - id: analyze-agent-transcript-skill-trigger-performance title: "Analyze Skill Trigger Performance" path: agent-transcript-analysis/3-analyze/analyze-skills/analyze-agent-transcript-skill-trigger-performance url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/3-analyze/analyze-skills/analyze-agent-transcript-skill-trigger-performance/SKILL.md tier: '3' description: "Per-Segment analyzer for false positives (Skills that fired but shouldn't have) and false negatives (Skills that should have fired but didn't). Recommends modify-description changes." references: - philosophy-on-skills - transcript-segment - id: analyze-agent-transcript-skill-action-performance title: "Analyze Skill Action Performance" path: agent-transcript-analysis/3-analyze/analyze-skills/analyze-agent-transcript-skill-action-performance url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/3-analyze/analyze-skills/analyze-agent-transcript-skill-action-performance/SKILL.md tier: '3' description: "Per-Segment analyzer for Skills that did run. Assesses helpfulness, token cost, and whether the Skill closed the loop. Recommends body refinements or deletions." references: - philosophy-on-skills - transcript-segment - id: analyze-agent-transcript-skill-gaps title: "Analyze Skill Gaps" path: agent-transcript-analysis/3-analyze/analyze-skills/analyze-agent-transcript-skill-gaps url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/3-analyze/analyze-skills/analyze-agent-transcript-skill-gaps/SKILL.md tier: '3' description: "Per-Segment analyzer for missing Skills. Identifies moments where a well-placed Skill would have prevented a wrong turn or saved turns, and proposes new Skills with a description and body sketch." references: - philosophy-on-skills - transcript-segment - id: analyze-agent-transcript-mcp-trigger-performance title: "Analyze MCP Trigger Performance" path: agent-transcript-analysis/3-analyze/analyze-mcp/analyze-agent-transcript-mcp-trigger-performance url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/3-analyze/analyze-mcp/analyze-agent-transcript-mcp-trigger-performance/SKILL.md tier: '3' description: "Per-Segment analyzer for MCP tools that fired when they shouldn't have, and tools that should have fired but didn't (e.g. agent reached for a CLI instead). Recommends description or surface changes." references: - philosophy-on-mcp - transcript-segment - id: analyze-agent-transcript-mcp-action-performance title: "Analyze MCP Action Performance" path: agent-transcript-analysis/3-analyze/analyze-mcp/analyze-agent-transcript-mcp-action-performance url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/3-analyze/analyze-mcp/analyze-agent-transcript-mcp-action-performance/SKILL.md tier: '3' description: "Per-Segment analyzer for MCP tool calls that happened. Assesses response shape, token cost, and whether the tool closed the loop. Recommends tool or server modifications." references: - philosophy-on-mcp - transcript-segment - id: analyze-agent-transcript-mcp-gaps title: "Analyze MCP Gaps" path: agent-transcript-analysis/3-analyze/analyze-mcp/analyze-agent-transcript-mcp-gaps url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/3-analyze/analyze-mcp/analyze-agent-transcript-mcp-gaps/SKILL.md tier: '3' description: "Per-Segment analyzer for missing MCP servers / tools. Identifies foreseeable closed-loop limitations the agent could not work around, and proposes new servers or tools." references: - philosophy-on-mcp - transcript-segment - id: analyze-cross-agent-transcript-patterns title: "Analyze Cross-Transcript Patterns" path: agent-transcript-analysis/3-analyze/analyze-cross-transcript/analyze-cross-agent-transcript-patterns url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/3-analyze/analyze-cross-transcript/analyze-cross-agent-transcript-patterns/SKILL.md tier: '3' description: "Cross-cutting analyzer. Given the per-transcript findings.*.json sets of several already-analyzed transcripts (the tier-3 outputs of analyze-agent-transcript), surface patterns no single transcript reveals: hindsight-as-foresight Segment patterns, recurring user prompts, deduped Skill/MCP gaps, and time-spend patterns. Still tier-3 labeling, but runs once over the whole batch \u2014 last in tier 3, an optional pre-report augmentation \u2014 not interleaved per transcript and not fanned out by the orchestrator. Writes findings.cross-transcript.json into the batch_dir for synthesize-agent-transcript-analysis-report to pick up." references: - philosophy-on-skills - philosophy-on-mcp - philosophy-on-prompting - transcript-segment - id: synthesize-agent-transcript-analysis-report title: "Synthesize Report" path: agent-transcript-analysis/4-report/synthesize-agent-transcript-analysis-report url: https://github.com/pulsemcp/ai-artifacts/blob/main/skills/agent-transcript-analysis/4-report/synthesize-agent-transcript-analysis-report/SKILL.md tier: '4' description: "Tier-4 synthesis. Runs once over a whole batch of analyzed transcripts: reads every transcript's tier-3 findings (findings..json from each per-transcript tmp_dir), plus findings.cross-transcript.json from the batch_dir when present, and synthesizes them into one consolidated report of actionable next steps across human prompting, Skills, and MCP servers. Writes findings.report.json (the recommendation slate) and report.md (the human-readable report, with the distance-from-ideal north-star block aggregated across the batch) into the batch_dir. Single-mode \u2014 a batch of one transcript is valid input, but there is no per-transcript mode; never driven by analyze-agent-transcript." references: - philosophy-on-skills - philosophy-on-mcp - philosophy-on-prompting - transcript-segment tiers: '1': acquire — get a transcript onto disk, redacted, in OpenTranscripts form. '2': decompose — build the recursive Transcript Segment tree. '3': analyze — per-segment analyzers across prompts, Skills, MCP and outcomes, plus a cross-transcript pass. '4': report — one consolidated recommendation slate across the whole batch. api_evangelist_authored_skills: [] api_evangelist_note: >- No skills were generated on PulseMCP's behalf. Generation requires an OpenAPI to ground operationIds in, and PulseMCP publishes none — inventing steps against the five documented REST paths would be fabrication. The MCP tool surface that DOES exist is captured in mcp/pulsemcp-tool-crosswalk.yml instead.