scholarseed-product-operations
Skill44 starsChanged 5 days agoApache-2.0View on GitHub
Access project resources and stored papers, PDF evidence and external retrieval, referenced conversation history, user-facing file delivery, explicit session-file imports into resource or literature libraries, multi-stage conversation plans, and MCP server configuration. Read only the relevant operation guide before acting. For file-only transfer/processing, read this skill and references/artifacts.md together, not evidence/search/library guides. Not for ordinary chat or self-contained answers that need none of these capabilities; research alone is not a trigger.
In this repository12 other files it gives its agents.
- academic-research-guidanceSkill
- agentic-searchSkill
- design-research-validationSkill
- fast-searchSkill
- find-skillSkill
- literature-reviewSkill
- pdf-evidenceSkill
- plugin-createSkill
- precise-literature-searchSkill
- project-literatureSkill
- research-idea-partnerSkill
- web-searchSkill
The file
We split this skill into the rules it gives an agent
10 rules. About 0 are common, 6 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 23
Use this skill to work with project materials, saved papers, conversation history, file delivery, plans, or MCP settings.
- Very specific: line 67
If you used a filesystem locator, obtain the file's URL with resource/stat.
- Very specific: line 68
If the question needs only a stored abstract, use literature/search then literature/detail instead of PDF list/read.
- Distinctive: line 71
Before the final answer, check that each used source has its returned navigation link.
- Distinctive: line 92
If discovery or publication fails, report that concrete failure and leave the files as unpublished working files.
- Distinctive: line 95
Keep file deliverables proportional to the requested task.
- Very specific: line 102
Use the scholarseed_bridge dynamic tool for product operations.
- Very specific: line 110
Before the first call to a product operation, obtain its request contract through scholarseed_bridge with …
- Distinctive: line 120
Follow the specialized skill named by the relevant guide when available.
- Distinctive: line 126
Do not route back to this skill as a research deliverable or force a product operation solely because a task is research.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Materials and Task Tools — 资料与任务工具
Use this skill to work with project materials, saved papers, conversation history, file delivery, plans, or MCP settings. Do not load every reference: choose only the relevant rows below, before accessing materials or starting the operation. A new need later in the task loads its reference then, not all remaining references.
Choose the requested operation
File bytes are not research evidence. Downloading supplied links, copying, renaming or converting files does not by itself require interpreting their content, discovering sources or managing a library. For file-only work, read this skill and the file-artifacts guide together and execute the requested operation; do not load retrieval, resource or plan guides merely because files are papers or PDFs. Do not explore product operations to find a more elaborate workflow for an already concrete task.
Load pdf-evidence when the requested answer needs PDF content, web-search when
external evidence discovery or verification is needed, and project-literature when
the task actually uses stored literature. Follow the corresponding guide then; mixed
tasks load only the guides needed by their requested parts. File acquisition does not
authorize library imports or bypass current-turn restrictions.
For tasks that may depend on existing materials, prioritize sources relevant to the question and goal: materials explicitly referenced or specified by the user first, then related files in the current session, then the current project's resource and associated literature libraries. Discover and retrieve project materials as needed without requiring the user to attach each item. Read only relevant content, not the whole library, and prefer internal evidence that actually supports the answer. This does not require visiting every source location on every turn. If internal evidence is insufficient, explain the gap before supplementing it through available external retrieval under the current turn's permissions. Provide references that lead back to the sources used; never present bibliography metadata or abstracts as full-text evidence. Do not search other projects or scan all local files.
Answering from project sources
Cite only sources relevant to the requested claim, not unrelated materials merely inspected during discovery. If no relevant internal evidence is found, state the gap without a source list or links to unrelated files or papers. Do not add an inventory of checked materials just to provide references.
For each relevant project source used in the answer, include a Markdown link
[source name](source_url), copying the returned source_url unchanged.
resource/read and resource/stat return file navigation URLs;
literature/detail returns the bibliography navigation URL and stored abstract.
A plain path, filename, or literature:<id> is not a clickable source link.
If you used a filesystem locator, obtain the file's URL with resource/stat.
If the question needs only a stored abstract, use literature/search then
literature/detail instead of PDF list/read; a missing PDF does not mean
the abstract is missing. Do not scan private state or the project tree to recover it.
Before the final answer, check that each used source has its returned navigation
link; if unavailable, state the limitation rather than calling a path clickable.
Navigation links do not establish PDF full-text evidence.
| Need | Read before acting |
| Discover, read, count, or modify project resources; use stored papers or notes | Project resources |
| Interpret a sent/selected/project PDF or retrieve external evidence | Evidence and retrieval |
| Answer from earlier or explicitly attached conversations | Conversation history |
| Create or deliver a file, or import a session file into a resource or literature library | File artifacts |
| Carry out multi-stage work, or continue/close an existing product plan | Conversation plans |
| Inspect, add, remove, enable, disable, or change an MCP tool server | MCP configuration |
Product Capabilities
For a requested file deliverable, read the file-artifacts guide and describe
artifact/publish before writing the files. After verification, publish each
requested deliverable with its intended role and inspect the successful receipt
before claiming delivery. A Markdown file link or a successful shell write is not
a publication receipt. Do not claim the bridge is unavailable merely because you
have not called it: if the tool is exposed, check its contract and actual result.
If discovery or publication fails, report that concrete failure and leave the
files as unpublished working files; do not invent a registration or a receipt.
Keep file deliverables proportional to the requested task. Once they satisfy the specified inputs and behavior, verify them, publish them, and finish. Do not delay delivery to add unrequested discovery heuristics, fallback paths, frameworks, or speculative refinements; expand only to fix an observed failure of the requested behavior. An independently reviewable stage deliverable can be verified and published when ready rather than waiting for optional work on the final files.
Use the scholarseed_bridge dynamic tool for product operations. Pass an object with
capability, operation, and request; for example, history listing uses
{"capability":"history","operation":"list","request":{}}. It is a model tool,
not a terminal command. Never invoke a similarly named executable, read bridge
credentials from the environment, or put a product operation in a shell script.
Skill examples that name a capability, operation and JSON request describe this
same tool call; the JSON is the request field, not a CLI argument.
Before the first call to a product operation, obtain its request contract through
scholarseed_bridge with {"capability":"contract","operation":"describe","request":{"target_capability":"<capability>","target_operation":"<operation>"}}.
Reuse that operation's contract only while it is visible in the current model context;
after context loss or compaction, describe it again. Fetch only the operation needed
now, not every contract. Examples below illustrate intent, not a complete parameter
schema; use the returned contract and real result identifiers, never guessed fields.
Contract discovery does not grant permission or replace any approval below.
Ownership and scope
Follow the specialized skill named by the relevant guide when available. Those skills own their evidence processing and business workflows; this skill owns only the product access, plan, and delivery contracts. It does not authorize downloads, publishing unrelated files, or other external actions on its own.
Research interpretation remains in academic-research-guidance, not here.
Do not route back to this skill as a research deliverable or force a product
operation solely because a task is research. Simple chat and direct answers
need no product workflow. If a needed skill or tool is unavailable, state the
limitation; do not invent calls, paths, results, or a substitute protocol.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

