Skip to content

scriptedalchemy/tracedecay

v0.0.0MIT

TraceDecay semantic code intelligence: daemon-owned code graph MCP server, agent skills, and a ChatGPT code explorer app with exact repository, worktree, commit, and generation provenance.

assessing-impact

Use when the user asks what breaks, what to re-test, or which tests cover code before changing a function, signature, type, field, or file. Load this skill before the first tool call.

code-health

Use when the user asks for a quality, architecture, coupling, or complexity read on a project or directory, or where its structural risks are. Load this skill before the first tool call.

diagnosing-analytics

Use when the user asks how TraceDecay is used: tool or skill usage, hook capture, attribution, or adoption analytics. Load this skill before the first tool call.

discovering-tracedecay

Use when you know the task but not which TraceDecay tool or operation covers it, including memory and administration tasks. Skip when routing-tracedecay already names a call for the question. Load this skill before the first tool call.

editing-safely

Use when the user asks to rename, move, extract, replace, or insert code across symbols or files. Skip when the edit stays inside one function body and keeps its signature. Load this skill before the first tool call.

exploring-code

Use when the user asks how a feature or flow works, where something is implemented, or to walk through code end to end. Skip when the user names the file to read or the text to find. Load this skill before the first tool call.

fixing-build-and-type-errors

Use when a build, compile, or type check fails and the error spans several symbols or files. Skip when one compiler message already names the line to fix. Load this skill before the first tool call.

inspecting-managed-skills

Use when the user asks about TraceDecay-managed skill proposals, activation, run evidence, or Hermes bridge health. Load this skill before the first tool call.

investigating-unexpected-changes

Use when the user finds working-tree, branch, or commit changes they did not expect and asks who made them or why, while preserving other agents' work. Load this skill before the first tool call.

managing-session-context

Use when the user asks what an earlier conversation or session said, or to recover prior-session context by time or Git history. Skip when the user asks for a recorded decision or durable fact. Load this skill before the first tool call.

managing-work

Use when the user asks to create or control TraceDecay Work tasks, attempts, graph mutations, placements, or adjudication. Load this skill before the first tool call.

managing-workflows

Use when the user asks to register, validate, activate, or retire a TraceDecay Workflow, issue a handoff, or control a Workflow run. Load this skill before the first tool call.

project-memory

Use when the user asks what was decided, agreed, or recorded about the project, or asks to remember a decision. Skip when the user asks what an earlier conversation said. Load this skill before the first tool call.

reviewing-changes

Use when the user asks to review, summarize, or assess the risk of a PR, branch, or working-tree diff. Load this skill before the first tool call.

routing-tracedecay

Use when you know the kind of code question but not which TraceDecay tool answers it; names the one call to make. Skip when the task administers TraceDecay itself (projects, configuration, storage). Load this skill before the first tool call.

tracing-functions

Use when the user asks who calls a function, what it calls, or how one function reaches another. Skip when the user wants occurrences of a string rather than call relationships. Load this skill before the first tool call.

using-the-cli

Use when TraceDecay MCP tools are unavailable or failing, or a call must run from the shell with the same arguments and scope. Skip when the daemon itself is down, since the CLI needs it too. Load this skill before the first tool call.